NEW CLUSTER STUDY THREAD!

[Helm]

Alright, this time for something a bit different.

I think I've honed in on the best possible challenge to learn good pixel art regardless of the style you want to employ.

I want you to pixel something (or take something you've pixelled before and fix it) and to make sure there are no stranded pixels in the image.

What do I mean by that? I mean that if there's a pixel of a specific color in your image, then there should be at least one adjascent pixel of the same color directly connected to it. I will accept 45 degree connections, but they should not be your only tool in solving the situations they come up.

I strongly suggest a small palette because you'll lose your mind if you try this directly with something that has 32 colors or whatever.  And I strongly suggest trying to represent a fairly realistic character or item to get the most out of this excercise. Be brave.

As usual, avoiding banding of any type is your primary priority. It's very easy to solve any stranded pixel situation if you're just free to band everywhere, after all.

Today at work I created a piece with very few single pixels and it suddenly clicked, for me. I've been pixelling for close to 20 years and only now it clicked for me, it's kind of ridiculous. Please try it.


P.S. I will accept single pixels only if you can explain the reasoning behind using them (and "it was too hard to find another solution!" isn't a good explanation).  For example, a specular highlight can not always be solved as 2x2 pixel or a 2x1 pixel. In those cases, use them, but they should be very few. Don't let yourself off the hook easily.

I don't have any art that I can show you because I pixel professionally for 8 hours a day now but if you're really serious about bettering yourself, give it a shot.

We'll discuss your findings as you post them.

[r1k]

glad I decided to check pixelation before I went back to work on my commissions.  Ill try to apply this practice to my work right now, though I wont be posting the result since its part of my work.  Anyways Ill look foreward to seeing the discussion in this thread.

EDIT: ok after cleaning up 2 frames of my sprite, the areas I find difficult to do this are outlines, and AA.  Outlines because of areas with 45 degree angles.  Even though 45 degree connections where allowed, but I tried to find other solutions.  And AA where sometimes you just want 1 pixel of AA like near a corner.

[Vagrant]

What?

But isn't this impossible? What about 2:2 pixel line Anti alias? Are we allowed to do that?



Challenge accepted, nevertheless.

[Helm]

If you want to anti-alias, better do it at an angle where two pixels or more would make sense!

You will never feel your clusters burning as much as when you simply can't do one-point antialias to fix every little thing by implying subpixels.

edit: about outlines, you can use 45s but do consider a 'jagged' outline, especially when you round a curve. I used to hate jaggy outlines (like, full staircase step jaggies) before, but they seem like such stronger clusters now instead of straight up outlines all over. But yes, for outlines it's an acceptable choice to use 45 degree connections - as I said I want this excercise to be style-transparent and I can't take one-pixel-outlines away from people and claim that it is.

Just keep it in mind generally: a single-pixel outline is banding by itself.

edit edit: just to clarify so the above statement doesn't seem like crazy-person-talk, a single pixel outline is banding only in implication, via negative space. It takes a trained eye to look at pixels like that and I'm not saying anyone will care.



I'm just saying, it's worth thinking about.

[Vagrant]

Professor Helm

(New)

(Original)


Here's an animated comparison gif.




...And here is where I couldn't help but break the rules. 



I might have missed some stray pixels here and there, but that's not a problem as I can go back and fix them. The ones in red are the ones I left there intentionally, because even tweaking a single pixel around facial features can radically change the original 'feeling' of a face; and I didn't see any alternatives.

This thing really made me think. It feels like a puzzle game. I had to get creative on the right cheek and even add hair there so I could lawfully anti-alias that line, and say GTFO to the jaggies there.

[Cyangmou]

Helm,
Some questions regarding to your working specs

I mean there are dither gradients which feature single pixels for the gradient. I am not only talking about the 50:50 checkerboard pattern, but also various other dither style feature single pixels as part of the well-sorted pattern.

Also how will this technique work stylewise if you are using shaded 45° lines where each single pixel will get one shade darker (bits and pieces of the chainmail)

I mean, I understand what you are trying to bring across, however I think what you are saying is heavily limiting you, especially if you are intentionally try to use banding and blurring to your advantage (the given example uses blurring gently outside the focus and far in the back to improve the depth) and as well as strongly used single pixel detail can get a necessarity for realistic textures here and there.
 
I'd say the more bigger, realistic and color-amount intensive you go with your style the less single pixels are mattering - there it's enough to get your big clusters right to show that you use clusters for defining your planes.
For realism or hyperrealism although you sometimes have to break good cluster theory in order to get effects working. Fully detailled realisms are to a certain degree "gritty"

For gamesprites and for simplified or "cartoony-colored" or "restricted-value-modeled" stuff - I underline your statement.

However I am not to fond that it works with every style, since you are explaining in my eyes a style-brick on it's own. 2:1 specular highlights also could be an interesting style decision on their own.


single pixels are everwhere
-mothholes in the cape
-texture gradients in the tree texture of the midground
-rivets (2x2 ones feature 4 "banding-aligned" signle pixels
-at the tail of flowing dither clusters (single 50/50 dither pixels in the water and the foreground tree bark)
-blurred half-sunken dirt stuff in the water
-small detail subpixel areas for detailling out lower res (nose)
and as well some not fully polished or just forgotten pixels as well - but I think the example serves it's purpose and shows up some interesting points.


I think it's debatable if heavily restricted cluster usage makes your pixels better.
As long as you use your clusters as value and plane definition I think you nailed the purpose of using them.

For every a bit more complex shaded piece it will get extremely difficult and time-consuming to get rid of every type of banding - and most-probably people who haven't calibrated their eyes to pixel technique most probably won't even be able to tell the differences.

I mean it's an interesting approach, but if you could provide some examples it'd help a leot for understanding you ( I am not 100% sure if I got everything you wanted to share)

[Helm]

Vagrant, there are still so many single pixels in that image! Do a second pass. I can spot dozens just by looking at it for a few seconds. That's part of the excercise, to look at it and train yourself to see the pixels.

Cyangmou,

1. yes, I am saying to avoid any dithering that creates single pixels. If one has a reason they absolutely need a dither pattern, they should explain what it is
2. if you want to do chainmail and can't find a way around it, do chainmail like that, but I bet there are other solutions to that textural problem if one tries hard enough
3. I do not believe banding can be used to one's advantage. Blurring will necessarily create *some* banding, but it should be minimized. I think it's completely within the scope of the excercise to do a blurred background like in your (awesome) image and have no banding or single pixels. It's just hard.

I am not saying everybody should do this for every piece of pixel art they produce. I am saying it's a worthwhile excercise

4.  I'd say the more bigger, realistic and color-amount intensive you go with your style the less single pixels are mattering - that is correct. And the less the piece benefits from being made with pixel art technique on the whole. On big, demosceney type pictures, I would expect the artist to show their pixelling in a few areas and do other areas much more vaguely and cgi-like.

5. I take your clarification in stride, even if we can't loudly proclaim that this excercise works for every style, it's still a valuable one to do, I think.

As to your image,  I wish I didn't have a dayjob, I'd sit and solve most of it and then we could discuss what is gained/lost by reshuffling the clusters. Select at least part of the image and try it, if you get the time. Don't just tell me why it can't be done, try to do it.

6. "I think it's debatable if heavily restricted cluster usage makes your pixels better.
As long as you use your clusters as value and plane definition I think you nailed the purpose of using them."

I think it's one thing to do that, and another to spend some time solving the 'puzzle', as Vagrant put it above. It gets you to a different mindset.

7. "For every a bit more complex shaded piece it will get extremely difficult and time-consuming to get rid of every type of banding - and most-probably people who haven't calibrated their eyes to pixel technique most probably won't even be able to tell the differences."

As with any involved finish in pixel art, the effect would be subconscious on the viewer.

I'll try to show something after work

[Cyangmou]

Helm,

I am not saying everybody should do this for every piece of pixel art they produce. I am saying it's a worthwhile excercise

I agree with that. The ultimative question is however what we gain from using it, where it's more powerful and where it just sucks in terms of efficiency. I am interested in finding that out.

1. yes, I am saying to avoid any dithering that creates single pixels. If one has a reason they absolutely need a dither pattern, they should explain what it is

I love to use dither pattern for various things
amongst others:
-soft light illustration (low-mid contrast diher patterns, just to show another kind of light)
-high res depth of field (low-mid contrast)
-low contrast gradient smoothing (low contrast)
-illustrating gritty textures (sand, gravel, other stuff depending on distance) (mid-high contrast)
-detail with a different color on a plane (rust, moss, ...)
-for blending clusters on top of clusters together

some of those effects can't be achieved with "blobby non-banding" cluster styles. But that's a choice you usually do if you start with adding dither.

I don't think that single pixels are bad or dither patterns, for me the key isn't cluster sharpness, it's the contrast. Low contrast single pixels don't hurt the eye.
On my new IPS Panel screen you can see the single of a low contrast dither gradient if you heavily focus those, otherwise it will "feel" like any other gradient.
There is the certain point in contrast where the effect tilts. But you can also tilt it effectively in order to show it's pixel art.

3. I do not believe banding can be used to one's advantage. Blurring will necessarily create *some* banding, but it should be minimized. I think it's completely within the scope of the excercise to do a blurred background like in your (awesome) image and have no banding or single pixels. It's just hard.

There we disagree. Blurring can definitely have effects of how to improve your image - lots of it is simple not worthwhile to do with pixel art or isn't achievable with low-res pixel art. But there are ways to use it, or at least I tried it and I think it's an valid alternative.
For a pixel artist the effect of blurring won't work, because he searches initially for clusters, for any other person looking at the image it will subconsciously be connected to an impression.
 
Minimize the"banding" never hurts.

It's definitely possible to do the blurred background in my image without any single pixels. The easiest way to achieve the same effect is simple with more colors.
I went there with dither and lower color to emphasize that it's pixel art.
Pixel artists always try to keep their palettes small, for better control over visual effects. But I think you have to choose between those 3 if you want to have fake gradients in your style:
-color count
-cluster sharpness
-dither

If you don't intend to have gradients the question won't occur anyways.

5. I take your clarification in stride, even if we can't loudly proclaim that this excercise works for every style, it's still a valuable one to do, I think.

As to your image,  I wish I didn't have a dayjob, I'd sit and solve most of it and then we could discuss what is gained/lost by reshuffling the clusters. Select at least part of the image and try it, if you get the time. Don't just tell me why it can't be done, try to do it.

The exercise is definitely valuable, no doubt.

I also don't claim that my example is the perfect solution regarding to style, cluster usage and how I solved various problems, but at least there are some underlying patterns and I can at least state what I tried to achieve in every place with various techniques.
It's already old, I moved on and would solve some things differently.
I am thinking a lot about how to use pixels to your advantage - a discussion in this direction could get interesting.

6. I think it's one thing to do that, and another to spend some time solving the 'puzzle', as Vagrant put it above. It gets you to a different mindset.

Knowing how to use clusters is a basic technique for pixel art, if not the most important technique and everyone who does pixel art should think about it during his creation process.
However you can band and break rules in art to your advantage and create different rules this way.

I mean I also break perspective or in order to get a more beautiful looking cluster or a "clean" line, but you can also do it the other way. Depends on what someones eye is calibrated to.
The pixel artist says: use clean line
The illustrator says: correct perspective is more important


Looking -forward to see what you are showing.
For this years Secret Santa I have made a piece with different focal style choices - we can discuss that one later too if you are interested.

[Vagrant]

Helm,

I am not saying everybody should do this for every piece of pixel art they produce. I am saying it's a worthwhile excercise

I agree with that. The ultimative question is however what we gain from using it, where it's more powerful and where it just sucks in terms of efficiency. I am interested in finding that out.

I agree. It is a tad time consuming -having it done myself-, and the only noticeable visual difference I've seen is that I have now more jags, and more limitations regarding small complicated areas that use multiple colors, like eyes.
Not to mention any line that is not a 'perfect line' (1:0 straight, 1:1 diagonal); you have to go really flexible and change plenty/re-work just to attempt to AA on anything that's 2:1 or god forbid, more.

Can't wait for an example.


Edit: I made a small 32x32 tree last night that I can apply this practice; see how it works.

[Lanarky]

I attempted to find all of the single pixels in Vagrant's image, and highlighted them in blue. I am pretty sure I found all of them, but I only spent around 5mins looking for them. I did this purely for the practice of spotting pixels that are out of place, and I figured that I'd post it (with permission) for other to learn from.

I can see how looking for them will let you navigate your work faster, but I am not sure how the final product will turn out quality wise, as it is hard to attempt without a concrete visual example of what we are trying to do by achieving this. Your post on banding and clusters really helped me advance my work, so I look forward to learning more from this. Thanks.

[Helm]

Tomorrow we wrap up on a game at work and I get time off for a couple of weeks. I'll try to push a personal image out of me (hard to do pixel art after pixel art, heh) to illustrate my point.

As to the main question:

"I agree with that. The ultimative question is however what we gain from using it, where it's more powerful and where it just sucks in terms of efficiency. I am interested in finding that out."

The main thing it would help beginning pixel artists with is to STOP them before they put down a single pixel. Make them consider it a brush stroke instead. A brush stroke has a direction.

Time and time again I arrive to some pixel-specific technique that turns out to be well-known in the larger theory of fine art. Not to say that the latter is a consonant space, there's various different schools of thought that are not compatible. But at least the concept exists somewhere there in the wider world of art that before you gesture at a canvas, you are considering the direction, weight and nuance of the gesture (of course the subconsciousness plays the larger part, but that's a nonfactor for what we're discussing here).

This is what this excercise is meant to illustrate. Pixel artists hold at their hand two amazing tools that other artists not often have, and almost never both of them at the same time:

1. minimization of physicality in drawing. You are not directed or constrained (or in the flipside, startled by) the subconscious drives of the body, as expressed through your moving hand on the canvas. So many people that are terrible with pencil and paper go on to become good pixel artists because they can plot down pixel after pixel, they do not rely on smooth movement from point A to point B

2. completely opaque, completely predetermined atom-level control of the picture. We sacrifice fidelity and in trade we get big, fat pixels that we can visualize before we even put down the first one. We can't visualize the whole image (difficult skill to train, real artists do train it) but we can certainly visualize a cluster, and with practice, a collection of touching clusters.

There are specific limitations and benefits to having these two elements in our art. Nearly all great pixel art shows telltale signs of how it was created (and that's another reason it's so trivial for a trained eye to spot reproductions/copies/photoshop jobs masquerading as pixel art). 

But beginning artists get lost in the art, they immediately begin SCRIBBLING. And then they salvage their scribbled thing with antialias.

This excercise is meant to stop people from scribbling. It's meant to teach them traditional art skills they'd get from proper tutoring that are specifically applicable to our discipline. A different path arriving to the same destination (which is, control, intentionality, visualization before action). Trad artists would go from general to specific. I struggled for years trying to apply general to specific in an organic way in pixel art and never could. I'd put down something vague, then zoom in and work on a corner of the image. Get my palette from that, and go systematically through everything using that invented palette.

It might have been alright and some of the art thus produced might be good in the eyes of others (or myself, sometimes) but that doesn't mean there wasn't a step missing.

Cluster theory was something I worked on to find a way to structure the middle part of creating a pixel art drawing. Not the heavy aa-finish, not the initial rough sketch, but the part where you're building the volumes and conveying the shapes. That's where clusters became a solution.

The excercise I'm suggesting here is a mental one. It's trying to help fix a step before clusters even occur. When the clusters are just in the mind. Ideally, one would consider the shape before they lay it down. But even if you go in and fix an older image after the fact, the mental training is there.


re: efficiency. The end result is less bad moves. Less pixels you have to fix. Less aa to take out later. Less blurriness introduced you then have to take out. Working with what pixel art has: sharpness, clarity of color, confluence of edges. Not working with what it hasn't: softness, transparency, vague shapes, smooth gradients.

I appreciate the capacity of an artist like Cyangmou that can put in pixels a blurry field of view effect. I do not say it has no place in pixel art (what does that even mean?) I am saying that for people that have gaps in the beginner and middle steps of creating something with pixels, they should consider not just bigger clusters, but their smallest ones. A single pixel is not a cluster.


As to what is gained from applying this technique: yes, the image will have harsher shifts, and will appear more blocky. Compromises will be made. But I think what is gained is a stronger gestalt, where the form and the content of the image are better aligned. A piece of craftsmanship that is declaring what it is, not playfully hiding what is is by showing it can emulate what it isn't too. To take a picture that has unpleasant blockiness but a promise of gestalt into a picture that retains the gestalt but has a pleasant blockiness is a matter of experience within that mindset. I am too, working on this.

[Helm]

I realize that we won't be getting started on this until I produce some results and damn it, I'm burning. I'm close to 30 years old, I've been pixelling for 20, seriously for over 15 and I am burning.


Do you see it?



If when you zoom in at a piece 'solved' in this way and it doesn't look like someone's playing tetris and losing, then it's not burning, not yet.

[Helm]

And here it is without any single pixel at all (so not even a single 45 degree connection). Perhaps even stronger, IMO. Bolder, smarter solutions with less pixel detail, more readable.

[r4c7]

While you were working on yours, I redid mine.
Original:

New:

Still have 45 degree single pixels, but removed all others. I also fixed the forced perspective and did some tweaks to the foliage and colors. (I actually got the color count down to 8 from 10) Still not happy with the colors, but I love the rest, even with all its quirks. Don't think I could fix some of the problems without adding more colors, so I'm leaving it. Really fun to work on.

[Ai]

This is an interesting exercise, so I took about half an hour and wrote a GraFX2 script to check it.

Here's a link to the script. Syntax-highlighted source code here

Installation:
* Paste the content linked above into a new lua file (mine is named 'clusteringcheck.lua')
* Save it somewhere that you'll be able to access easily from GrafX2's file browser.

Usage:
* Rightclick on the 'Brush effects/Factory' tool button (to the immediate left of the 'text' button)
* In the dialog that comes up, navigate to where you saved the script and select it.
* Click 'Run'. A dialog will appear with a single checkbox.
* Click the checkbox next to '8 way neighbourhood' to turn off 8-way neighbourhood if you want (Meaning that only 90-degree adjacencies will be considered; pixels with only 45-degree adjacency will be considered errors.)
* The script will set color 255 to bright red (RGB 255, 0, 0), and mark every 'bad pixel' with color 255.
* After it's done this, it will also popup a report on how many pixels were bad out of how many total.

Here's some example output:

Initial image:



4way check (45-degrees not allowed):



8way check (both 90-degrees and 45-degrees allowed):




I've also tried it on other, larger images, like Fruits.

Here's Vagrant's 'new' image:


The result of checking it with 4-way adjacency (7% of pixels were erroneous)


The result of checking it with 8-way adjacency (1% of pixels were erroneous)


Those of you who are GraFX2 users, enjoy!

[Helm]

[sneaky_elm_at_work] Ai, thank you for this, it's going to be very handy![/]

[Ai]

No worries, glad to see someone will get some use out of it.

Here's my first try at this exercise (from scratch):



Not sure about bandiness of bottom part, but I'm happy that it is completely solved (no error pixels)

That reminds me to mention: if no error-pixels were found, no report is generated. Perhaps I should change that, what would you prefer?

EDIT: I've updated the script to make a report in all cases.new script here.

EDIT2: Here's another try.
This one is a fix of an existing icon.

Original:


Fixed:


I find it a struggle not to mutilate the planes when I'm trying to fix clusters.

[ErekT]

Great exercise It felt very counter-intuitive to not be able to use any single pixels in the beginning, but it forced some new shading solutions superior to how I've approached shading before I think.

I used a portrait I drew a while ago and wasn't happy with at all, back when I thought that dithering in human faces was a good thing to do Still not happy with it, but improved in terms of pixel-tech at least.



->

Tried to get rid of all the single pixels but there's probably still a few on the loose here and there that I missed. Also I kept the single-pixel eyeglint on purpose. Couldn't get the eyes to work without it.

[Vagrant]

And here it is without any single pixel at all (so not even a single 45 degree connection). Perhaps even stronger, IMO. Bolder, smarter solutions with less pixel detail, more readable.

Oh, I see it now. Definitely. 

As to what is gained from applying this technique: yes, the image will have harsher shifts, and will appear more blocky. Compromises will be made. But I think what is gained is a stronger gestalt, where the form and the content of the image are better aligned. A piece of craftsmanship that is declaring what it is, not playfully hiding what is is by showing it can emulate what it isn't too. To take a picture that has unpleasant blockiness but a promise of gestalt into a picture that retains the gestalt but has a pleasant blockiness is a matter of experience within that mindset. I am too, working on this.

Yeah. But.. Stronger gestalt?

This is entirely subjective, and what I would call a pixel 'style'. I can't see this applying to any other forms of pixel styles (Like my piece before) without radically changing it into 'sexy cluster style', that overwrites the original. Using only clusters can limit you incredibly in portraying certain aspects, especially if they are small.

For the purposes of the exercise, it's great. But again, it's limiting everything to this clustery 'flavor'.



Here's a newer one that's true to sex cluster style. You can't get any more cluster-y than this.

[Ai]

^ That's a great example of what I mean by feeling forced to mangle planes. Normally certain depth information is conveyed by overlapping forms, which would tend to create occasional pixel-islands, and these rules really punish creasing -- the tighter the less doable.

[Helm]

Honestly, I think this is up to skill in execution. I could have edited the mouth and nose of your character without having changed the style and still have no single pixels. The reason I did (change the style) is because I wanted to maximize what I could do with that particular palette to convey something realistic, as I found it more challenging.

The pixel work I plan to do in this mode of no single pixels will not be cartoony, and it will be heavily planar, because that's jsut how I want my art to look. It's up to other people to try to prove me right, not wrong, by trying to do their own particular styles within this limitation. That's the only way to know for sure if what I suggest is style-independent.

[Cyangmou]

Honestly, I think this is up to skill in execution. I could have edited the mouth and nose of your character without having changed the style and still have no single pixels. The reason I did (change the style) is because I wanted to maximize what I could do with that particular palette to convey something realistic, as I found it more challenging.

The pixel work I plan to do in this mode of no single pixels will not be cartoony, and it will be heavily planar, because that's jsut how I want my art to look. It's up to other people to try to prove me right, not wrong, by trying to do their own particular styles within this limitation. That's the only way to know for sure if what I suggest is style-independent.

Yeah, planar lighting and construction has to be done right, that has nothing to do with how we lay in clusters. Vagrants new sprite just lacks form, but that has nothing to do with the technique helm mentioned.

After testing it out for my own (example will follow December 25th-27th)

"The ultimative question is however what we gain from using it, where it's more powerful and where it just sucks in terms of efficiency"

I'd say the technique could also be referred as "cluster based noise reduction"
It really brings out a lot of unecessary pixel you don't see unless you have 2 layers and kicking them out on one - or (most probably) you calibrated your eye to see this kind of thing.
This technique forces one to overthink how to form clusters

As to what is gained from applying this technique: yes, the image will have harsher shifts, and will appear more blocky. Compromises will be made. But I think what is gained is a stronger gestalt, where the form and the content of the image are better aligned. A piece of craftsmanship that is declaring what it is, not playfully hiding what is is by showing it can emulate what it isn't too. To take a picture that has unpleasant blockiness but a promise of gestalt into a picture that retains the gestalt but has a pleasant blockiness is a matter of experience within that mindset. I am too, working on this. 

Yeah if the technique is used for the whole image. However I think we should handle the whole thing like every other technique with a grain of salt and apply it only where it makes sense and were we want to have clear planes and sharp descripted forms. Like over-AA and over-dithe rthere is over-cbnr as well

Combining this technique with other techniques is possible.

I think for dither the technique can also be applied, however it needs another kind of ruleset because dither gradients feature 45° connected pixels. Easiest rule for this would be allowing 45° connected dither pixels and allowing single pixels for fading dither patterns

For single pixel specular highlights the technique won't work. We even shouldn't try to apply this technique there.

Especially for highly characteristical forms like facial features I suppose that subpixeling with single pixels is much more powerful, just because a human being knows exactly how a face looks like. Unless one uses the technique strictly for the whole piece I'd leave it out there.

I agree that for an aspiring pixel artist or a novice it's a valuable rule which should be taught in order pretend him from establishing bad manners and using bad gradients.
After all clean lines look sharper and nicer than "meaningless fake-dither" and forces him to build up an image with planes - even if the light construction on those planes is off the result still looks nice.

[Cyangmou]

as promised an example from my side



Short explanation what I wanted to do with the piece:

I tried to create sort of a realistic impression with a low-res canvas and fat and chunky pixels and as well as really low amount of colors - palette consists out of sweet 16.
Furthermore I tried within those harsh restrictions to capture a golden hour impression with a focus on chiaroscuro for an exterior scenery.
I used checkerboard dither for the illustration of soft lighting on nature which contrasts harshly with the clear cut light on the gothic architectural forms.
Single isolated pixels were mainly used for emphasizing some highlights and adding a tad more detail to the procession of knights.
I also played a lot with size relationships of single elements to increase the impression of the greatness.

Well, so the "CBNR" version shows the cleaned version with applied rules for cluster, AA and dither. The other version shows the version without extra polish (or overdone technique).

The CBNR version definitely looks a lot more polished, goes lighter on the eyes and looks sharper. Plus if I'd have done it right away with CBNR I most probably would have saved a lot of time.

However I am finally really questioning myself now what's good "pixel art" and what's good "art made with pixels". I am having a lot more questions now, than before.
I mean this simple rules lead to quite a big difference in the whole thinking process and this rules don't leave you another choice than ending up with really clean pixel art.
I mean one of the biggest choices I really had to do were I intentionally wanted to put a highlight and where I don't wanted to do it. THere are barely any "scribbled" or "unintentional set single pixels" left.
The group of knights shows this process of cluster polishment and highlights quite good. In this area there are also the most single pixels in the whole image, due to really small scale.

I mean the piece showcases quite well a completely different style with completely different qualities than the other realistic image from last year. Also in terms of size and palette it's more optimized to this technique than the other one.

Maybe someone wants to add some thoughts.

[r4c7]

On the topic of this technique, I feel it should be fleshed out more as an idea. Maybe add more rules in support of single pixels.
Here are my ideas, but as an amateur, take them with a grain of salt:
Pixel art tries to push the boundaries of bitmap images. Sometimes there will be a "shape" that is too small to be more than one or two pixels or too big to be ignored. This picture is an attempt to show how a complex shape (below) may be portrayed in single pixels (above). Sometimes this may force us to use single pixels. Sometimes it may be better to change the form of the shape to make a better cluster with less single pixels.
Antialiasis shouldn't be explained as making an outline smoother, but as a way to better portray a shape/cluster.

I don't know how to explain dithering. Just my 2 cents here.

[Vagrant]

I recently discovered thanks to this, that I tend to lean towards art made of pixels, rather than the purist 'pixel art', as proposed here. But not as extreme as a 'pixel drawing', made without any knowledge of the pixel polishing basics.

When I dive into the craft, it's almost always a matter of eyeing the preview box at x1 often for a great visual aspect at that scale, and I'll negotiate with as many techniques as I can until it looks right.

This is different though. It's a unique visual style that seems to look better when zoomed crispy without blurs. Very clean. 

[Arne]

The whole cluster approach reminds me of the Photoshop filter Noise>Median which blobbifies an image. I sometimes use it in my analog'ish digital art to get rid of unwanted noise. The effect smells a lot like filter though, so I don't like using it. Similarly,  I think a pixel art image which has been heavily clustered/noise reduced (and posterized) can be... rather boring to look at. All the interest points and subtle little scuffs, knobs and trinkets have been washed away. There are no depths to explore once you've taken in the immediate surface.

Also, I think clustering is more effective at a low color count (great difference in color) where haphazard noise really shouts for attention and you need to be graphically clear about everything (detail and subtle stuff can be quite unaffordable). In the case of the lady face which has so many colors, I think orphan elimination isn't as important. In the case of the castle image, I think some of the peninsula erosion is detrimental as it truncates the range of expression (makes things a little blobby).

[RAV]

If we go by the idea that every school of art is an actual... school, that is to teach something specific, instead of a genre definition that excludes everything else, then the regular training in the core tradition is to help you improve even in your divergence, rather than tell you not to.

[Helm]

However I am finally really questioning myself now what's good "pixel art" and what's good "art made with pixels". I am having a lot more questions now, than before.

I mean this simple rules lead to quite a big difference in the whole thinking process and this rules don't leave you another choice than ending up with really clean pixel art.

Exactly.

So, as a thought experiment, if we put this process at one end (actually, even moreso at the end of one spectrum we should put not-even-45-degre-connections-allowed) and at the other end we put demoscene art that is pixel-by-pixel (let's consider that there was no blur brush or whatever) but at such a high color count and so overwhelmingly dithered that a single pixel really alters nothing on its own by being or not being there, then we have a range of expression.

In that range of expression, when one sits down to draw something with pixels, they might want to consider where their piece (or parts of their piece?) will sit. And for what reason they chose where it'll sit.

That's all I'm interested in, as a method of learning, for people to consider the pros and cons before they start, even if they deviate after they start.

As to that feeling you got about what is good pixel art vs what is good art made of pixels, I think that's a valuable issue to explore further, just as long as we do not put a value judgement on either way. Let's focus on the pedagogical quality of learning 'good pixel art' or 'good art made of pixels'.

Antialiasis shouldn't be explained as making an outline smoother, but as a way to better portray a shape/cluster.

Either can be valuable though I would choose to go with the latter, but what if there's a beauty to just using clusters that do not need to be smoothed out to suggest something else that isn't there exactly? That's the point of this excercise, it's where the form informs the content, and not only vice versa.

Vagrant, I hear you, and that's fine. I just would suggest to not make any final decisions about this stuff just after considering it for a week. I used to AA just looking at the x1 zoom preview as well and that was when I was over-aaing the most and creating subtle banding and then I'd spend so much time trying to take it out. What if I hadn't put the over-aa in there to begin with?

As to whether the method I'm suggesting looks better zoomed in or zoomed out, I stand firmly in the side that says it looks crisp and beautiful at 1x and like a dizzying game of tetris up close (which, if I'm not being clear, to me looks like a good thing). If something just looks better at 1x zoom but falls apart close up, there's an element of misdirection in there somewhere. Or perhaps that's just how I feel about it, let's not state it like it's objective.

Arne, I absolutely agree the benefits of this are most apparent when you don't have too many colors.

Tomorrow I'll be starting my own piece using this technique, and at 90 degree angles as well. There will not be a single single pixel.

[0xDB]

Based on...
"A group of pixels (more than 1 pixel) of the same color, making a shape."
...and...
"No 45 degree connections allowed."
...
I decided to start my exploration of clusters in a formal way, by thinking about how those clusters would form from the void.

So, I took a single pixel (not a cluster by definition) and started finding all possible variations of it by applying mutations (defining a mutation as an addition of exactly 1 pixel to it, according to the "no 45 degree connection rule"). It quickly became apparant, that down the generation tree, there would be many repetitions of clusters, as results of different mutations. Interestingly, on the third generation(or fourth if you start counting at 1 instead of 0), all classic Tetris blocks were born (see image below).

When I saw that I felt this could not be coincidence, so I did a little searching and found there is already a mathematical name for what we are dealing with here. A pixel "cluster" without any 45 degree connections (and assuming the pixels are squares) is called a Polyomino (http://en.wikipedia.org/wiki/Polyomino) (and the Tetris blocks are all Tetrominos).

So, I suggest the use of Polyomino as a term when referring to those types of clusters for it appears to be a well known term which has been around since 1953.

[RAV]

Though, if we go by the notion of pixel art as a conscious mode of work to learn about, rather than definition of happenstance result, then it becomes a "platform agnostic" principle. That means, the truest teaching of this school is universal in application as much as recognizable.

The pixel art is first and foremost based on a morphological idea of pixel. The pixel changed over time, as the physical definition of hardware changed, both in possible colour and dimension. Thus it can also be abstracted to a "virtual pixel", which in its most base form is simply a scaled pixel to emulate its history.

But why stop there. Part of my own research on virtual screenspace is to show how weird it can get, while keeping all artistic lore applicable:


Then of course, most should already know when I did this:


Most importantly though, what you might not yet know is my abstraction to it:


So, looking at the end of the last, what if a single "pixel" is not only 3d, but also of arbitrary size and even in form of a "half-pixel", which then is a natural 45°.

You might feel I am stretching it now, and misappropriating this thread, and maybe I am. But if nothing else, at least cluster theory is relevant to it, so this is where my interest comes form.

[RAV]

The virtual screen is within a physical host screen, and thus a virtual pixel can consist of many more physical pixels. So what I mean is, we now have this interesting constellation of understanding, two kinds of 45°, and my interpretation of 45° does not exclude yours, because the virtual level pixel 45° is actually made out of 90° physical level pixels. But on increasingly HD displays, the physical pixel becomes increasingly meaningless to pixel art, what remains relevant is only virtual. So what you see below now will appear tiny on screens tomorrow.

Compared to what you discuss, what my engine does is this, cutting pixels in half:


It might sound absurd to bring this up, but the reason I mention this is because the virtual pixel is handled in artistic understanding and workflow the same a real pixel, you even very much depend on that today, and I believe what I propose has quite a good effect without infringing on the basic principle of pixel art as layed out here.

[Ai]

^ Your voxel videos are interesting, but I have no idea what you propose at all.

As far as I can tell, the point of eschewing dither and AA is to avoid being mesmerized by technique, and make pixel art that is actually well constructed rather than pixel art that merely looks good. How you conceptualize pixels doesn't appear to effect that -- isolated voxels have a very similar visual effect to isolated pixels.

[RAV]

The point is that it has been brought up in debate that for practical intends and purposes, there is a balance to strike between "well constructed" and "looks the way I need". There are occasions isolated pixels are necessary, though this training here is supposed to make them a rare and conscious effort. Now the problem is that of all cases of isolated pixels needed, the 45° case is the most needed; it is not truly isolated, but not truly well constructed either.

To handle that case so it satisfies both, we need to rethink what a pixel is about really, its history and future, that it is not this fixed thing set in stone, demonstrated in my examples. Thus, arriving at the virtual half-pixel is to bridge that gap of 45°, and give this theory a practical polish.

However, in the end this is just my way of handling it, that I thought would warrant mentioning related here.

[0xDB]

I feel like I am doing it wrong:

The goal here was to polyomino-fy a small part of my avatar to get a feeling for clusters. It seems impossible (for me at least) to achieve the same level of detail/crispness without any 45-degree-connected pixel aggregations. The result appears blurry, mushed and dirty to me and I am incapable of reproducing the original shapes and angles in those limitations. The mushiness could be a problem with the palette and not as much caused by the polyominoes alone but the angles are an effect I attribute directly to the strict adherence to 90-degree-connections.

[Helm]

The problems you're having have to do with that you're constantly trying to *reduce* clusters,where sometimes you have to expand them, let them become more robust, have a stronger shape that reads better at 1x.

[0xDB]

I think I am beginning to see. Using your edit as a reference enabled me to make a second polyominofication which resembles the original much closer and it appears crisp too. Eliminating each and every 45 degree connection between any two same-colored pixels was just as tricky as trying to preserve the AA (which I am starting to think is something I should get rid of altogether as there is no such thing as AA in reality, just surfaces which can curve towards or away from the viewer/light at more or less steep angles, making smooth or sharp gradients).

[RAV]

Though I should have made it clearer maybe that my technique is not meant to fight the polyomino approach, it is not necessarily meant to encourage what you mean by 45°.

What it will most likely be used for, by way sculpting gives incentive, is making polyominoes, and then having them smoothed into any finer resolution on scaling, maintaining 90° integrity on finest (hardly visible) level -- that is what I mean with 45°: when it is scaled up rough, it's not quite as rough, but it's still polyominoes. On physical x1 both look the same.

That would be like taking Dennis' last result, scaling it up a lot (not zoom, actual), and then using the finest pencil size to add or cut the corners of those pixel-squares, just that a tool would do that automatically on runtime for any scale on the fly.

It is a kind of polish, optionally. Whether this is still within the nostalgic aesthetic you're looking for, if you prefer it rough in any case, or its own kind of "tiled" aesthetic, is another question. Maybe it's already a distinct style that just complies to this cluster theory on the lowest level. However, for my voxel case especially, I regard it more important, because the extra dimension can interfere more on perspective, and "scaling" is naturally dynamic in fps as well. So I mention this from the outlook of how cluster study translates.

[Cyangmou]

@RAV

I honestly dunno why you posted that kind of stuff in this thread here.
Your approach seems to me work more like certain algorythms to unpixel pixel art and adding "polish" to something. I don't like that kind of "polish" and I don't think it's needed nor does it make the impression better.

That the pixel changed over time and that it's rather a certain aesthetic nowadays instead of state of the art technique is also common knowledge.

As Helm already stated in the thread there are things pixel art is very well suited for and there are things for which pixel art isn't the best medium.

That's all I'm interested in, as a method of learning, for people to consider the pros and cons before they start, even if they deviate after they start.

After all a lot of the stuff I experimented with isn't "the best approach to pixel art" since another medium will be much more effective, but I am mainly doing that stuff to find out some limits of the medium.
The whole idea of the cluster approach is just to bring out the aspects for which pixel art is considered great while neglecting a lot of other aspects.

___
 
@Dennis:
Great link to the polyminos

___




The simple ruleset we are using here makes it quite easy to end up with clean and up-to-date pixel art.

One question for all the technique interested people here:

I just float the statement that AA is working against the ruleset
what's the general opinion on that?

Do you even believe that AA is necessary with modern hardware and pixel art graphics?
modern flatscreens are really sharp and there is no blurring effect, so AA also don't has the same effect as on an old crt monitor (same as gradients and bands led us to the whole cluster approach).
AA anyways don't works efficiently with barely any single pixel.
So is it better just to cut it out completely or does a common 2-4 wide AA buffer cluster serves it purpose?

The aesthetic without AA has more jagged lines, but the pixels are clearer visible as well.
With minimizing the colors and emphasizing the clusters we will end up with a simpler picture anyways and we don't even try to hide that it's made out of limited colors.

So basically using a buffer cluster should be the same decision as if we'd use a highlight.

[RAV]

The topic of the thread is Cluster study, not Pixel Study. So I took the liberty to interpret it in universal appliance. It's relevant to more than pixel, from cross-stitch to Pearler bead, to Rubicks to Tetris to Lego to Voxel; most techniques cross-pollinate, because there is a common logic to them. My participation here was not meant for ranking one aspect over the other, but contributing to the wealth of the topic. What I do is not a replacement to pixel art but an opportunity for a pixel artist to naturally translate his skill in the theory here into yet another field of craft, made more familiar when it otherwise might be too alien. And my explanations were a narrative towards that, why I do what I do, the differences and extensions born out of challenges I am facing in my case of application, that I thought would be interesting to you, and in that exchange of expertise I was curious to hear constructive opinions on that related matter from experienced artists. But I guess I'll just stop bothering.

[Helm]

re: translating this theory in other crafts, sure perhaps that's a useful thing to do but we've only started to get a grasp of pixel theory *as* pixel theory per se.


Cyangmou and Dennis are moving towards the same issue with AA under this ruleset.

My answer is: ....I do not know, yet. I agree that AA has changed as a necessary and valuable part of pixel technique as screens have gotten sharper. I've seen a lot of people go for no-AA works that I have found very aesthetically pleasing. And it can be said that a tour de force of super-AA smooth pixel art seems to devolve: it looks like vector art and one has to wonder why it isn't vector art, then.

However in my own work I used to have great trouble letting go of AA, as a lot of skill I've gotten in this art style was related to that. And there is certainly smart AA. And much of a good, mixed color palette is about reusing a main shade as a buffer shade somewhere else.

So through this new idea about minimizing stranded pixels, I finally have a systematic way of figuring out how much AA I really need, if any. I do not have any solid answers to your query, Cyangmou. Perhaps 2point AA (like a small flat cluster) will look to me as gimmicky in a few years as single pixel AA does now. Perhaps 2point will survive because it looks harmonious when it's not aligned with other edges of two-pixel clusters (when it's not banding, so to speak).

And another startling thing about this whole deal: even banding looks less bad when there's no single pixels around, for some reason. Somehow the tetris jumble allows for it more, not sure why.

[RAV]

I don't believe that all useful insight into cluster theory has originated first, and must originate in all things further, solely from what we call pixel art today, before we can move forward, or that there'd be no synergy in this quest, I believe much of your inspiration to form your thoughts on this has been interdisciplinary to begin with, but for now it seems difficult to align our special interests. Maybe some other day. I see you have much to think about as is, and so have I, neither of us depends on the other. I'll refrain from further deviance here.

[Cyangmou]

My answer is: ....I do not know, yet. I agree that AA has changed as a necessary and valuable part of pixel technique as screens have gotten sharper. I've seen a lot of people go for no-AA works that I have found very aesthetically pleasing. And it can be said that a tour de force of super-AA smooth pixel art seems to devolve: it looks like vector art and one has to wonder why it isn't vector art, then.

Yeah, there is the vectory pixel art around (e.g. Panda's work) and yeah I also doubt that this kind of style is optimal for pixel art.
And overly smooth AA also seems to wash away the sharpness we want to have emphasized.
I mean if a piece of art should have this kind of look, an artist would go and use a vector program and he won't have to deal with getting it smooth at all.
I mean in terms of efficience it's definitely a bad choice, you spend first a lot of time with creating the piece and then you have to smoothen out every single bit of it, if there are lines in it it gets even harder to make them consistent in thickness with AA
I mean for showcasing that you completely understood how good you are with AAing it's ok
the bad side about this is that there are lots of single pixels and it looks really blurred if you look at it at 200% or bigger

(example for all who don't know Panda)



vector or pixel at 100% ?

However in my own work I used to have great trouble letting go of AA, as a lot of skill I've gotten in this art style was related to that. And there is certainly smart AA. And much of a good, mixed color palette is about reusing a main shade as a buffer shade somewhere else.

yeah, I don't talked about good mixed palettes. It's possible to use each shade to it's maximum.
I mean if you do a technique over and over again and you just get used to it and it gets second nature to you it's really hard to let it go - especially if your sense for aesthetics is telling that there is something missing.
However the question is if we should see AA just as time eater and "bad habit" if we change our style aesthetics to the cluster approach.

So through this new idea about minimizing stranded pixels, I finally have a systematic way of figuring out how much AA I really need, if any. I do not have any solid answers to your query, Cyangmou. Perhaps 2point AA (like a small flat cluster) will look to me as gimmicky in a few years as single pixel AA does now. Perhaps 2point will survive because it looks harmonious when it's not aligned with other edges of two-pixel clusters (when it's not banding, so to speak).

I think if we leave out half part of the technique leaving it all out at won't really hurt. "gimmicky" describes it quite good.
I mean AA is one of the first techniques which a person who is new to pixel art learns and it needs time to fully understand it. Pixel artists just look for AA in pixel art pieces because they expect they would miss something if it'snot there.

Your systematic way would be great to know of - I mean I am usually solving it just with intention


But there are occasions where you might need single AA Pixels - the thing I am talking about are perfect circles - gets really hard to smoothen out a round edge without them. Or leving them out completely

Because I am lazy I am just using Cure's examples.
any need to "fix" it with a cluster approach?
Or just don't thinking of AA at all?

And another startling thing about this whole deal: even banding looks less bad when there's no single pixels around, for some reason. Somehow the tetris jumble allows for it more, not sure why.

90% of banding occurs if you have too many colors in too less space.
I think the tetris jumble allows more banding because if we start looking closer at it, the whole art looks tidy because of intentionally aligned forms. I suppose the cluster-form impression is then percepted stronger as if there are just thrown in pixels so to say.
I mean the polymino approach (if used strongly for detailled bits) lets the whole thing look more and more like a pattern if you look closer.

[Dex]

I completely understand the idea and concept behind doing an exercise like this (I actually went over my avatar and did a version that was void of single pixels, but alas my computer shut down and I hadn't saved it...) but to inherently change the way you pixel just to eliminate an abundance of single pixels? That's an entirely silly and futile idea in my opinion. Who are we to say what we should or should not do with pixel art? If someone wants to create pixel art that is smooth enough to be a vector image, why shouldn't they? Pixel art is a means to an end, just like any other artistic medium. You can create whatever you like with the medium, and saying "well, they simply could have just done a vector image instead" does not hold any water in my book. Sure, they could have, but in the end it is entirely up to them. It's not a waste of time or a worthless exercise just because what they made could have been made with another medium. People create oil paintings that look like photos-- should we tell them to stop painting or change their style? Of course not.

Elk's incredibly large and detailed dragon pixel art--- could that have been created digitally in Photoshop as opposed to pixels? Surely. It doesn't change the fact that his artwork was impressive, incredibly intricate, and well crafted. He didn't waste any time by not making it in Photoshop instead because it was his choice to utilize pixels in a way that he enjoyed controlling and working with.

The same goes for this concept-- the concept of clusters and "incredibly clean pixel art" is surely interesting and can indeed help us decide where we place our pixels in the future, but I have qualms with the idea that we should eliminate single pixels from the process or dismiss the ideas of dithering and AA for the sake of a more optimal sort of pixel art, since we can't clearly define that term. What the ideal pixel art image should be composed of is entirely opinion based.

Over time my pixeling style has worked to attempt to eliminate banding in any shape or form, and I attribute that to the studies done both here on Pixelation and how I've learned and grown over time as well. I know that the idea proposed here isn't to truly eliminate working with single pixels, but some of the talk that has arisen about AA not being a worthwhile exercise and stating that pixel art that is smooth enough to be vector art shouldn't have been pixeled at all just did not sit quite well with me. Lines like "the bad side about this is that there are lots of single pixels and it looks really blurred if you look at it at 200% or bigger"-- the "bad side?" How is that a bad side? It's entirely up to the viewer. Panda's image shows mastery of several pixel art concepts, and containing single pixels does not constitute a "bad side" in my opinion.

I don't think Helm's original intention with this topic was to propose changing the fundamentals of pixel art entirely, but some of the ideas and arguments that have arisen just seem contrary to the idea of pixel art being just that-- an art form composed of pixels. From my perspective, how we utilize these pixels should not be routinized nor categorized.

[r4c7]

I don't mean to drag this out into a new AA thread, but I whipped an example up. I'm sure what I'm going to say has been said before, but its been on my mind, so I'll let it out.

I find it interesting how the different contrast levels affect aa. in A with 100% contrast, aa is definitely needed. I also had  to use two aa colors to make it look right. B, with 50% contrast, only needed 1 aa color. C, 25% contrast, there is no aa and it looks fine. Its interesting how contrast effects the need for aa.

Then all of a sudden, I thought how these rules apply to dithering. (Top is dither, bottom mixed color. In C, all are dither)

With A, there is no room to use 2 extra colors to aa between the black and white, so it looks bad. B looks okay. C it is hardly noticeable there is a dither pattern.

It's kind of like dithering is a type anti-alias arranged in a tight repeating pattern, if that makes sense. I'd also be interested to see if it would be possible to make a dither pattern involving more than two colors, maybe even 4, to further support this. I may edit this a bit, or make a new post, about what I'm thinking, but I think I need to go to bed soon to refresh my brain.
I feel like this is borderline gibberish to the more experienced artists here.

[Ai]

I completely understand the idea and concept behind doing an exercise like this (I actually went over my avatar and did a version that was void of single pixels, but alas my computer shut down and I hadn't saved it...) but to inherently change the way you pixel just to eliminate an abundance of single pixels? That's an entirely silly and futile idea in my opinion. Who are we to say what we should or should not do with pixel art? If someone wants to create pixel art that is smooth enough to be a vector image, why shouldn't they? Pixel art is a means to an end, just like any other artistic medium. You can create whatever you like with the medium, and saying "well, they simply could have just done a vector image instead" does not hold any water in my book. Sure, they could have, but in the end it is entirely up to them. It's not a waste of time or a worthless exercise just because what they made could have been made with another medium. People create oil paintings that look like photos-- should we tell them to stop painting or change their style? Of course not.

Relax, this is not a thread about any kind of prescriptivism.

 Cyangmou is simply saying 'an efficient artist would not do this, because it's an inefficient means of achieving the effect.'. He's talking about opportunity cost: the fact that time spent on rote, repetitive parts of your image is time that you could have spent on the basic ideas and fundamental grounding of your image. The spirit of the image suffers as a result -- it quite naturally becomes more about technique than concept, because that is how the artist chose to allocate their time.

If you want to do a few images that are just about getting solid AA skills, more power to you, that may well be a smart use of time. But after mastering AA, you go on applying mass amounts of AA, that may not be a smart use of time. It may be a waste of your humanity in a very literal sense; choosing to put technique in when you instead could put a more meaningful part of yourself in.

Similarly, Helm is saying 'This exercise will make you think, try it.', not 'Exterminate all of the isolated pixels'.

Over time my pixeling style has worked to attempt to eliminate banding in any shape or form, and I attribute that to the studies done both here on Pixelation and how I've learned and grown over time as well. I know that the idea proposed here isn't to truly eliminate working with single pixels, but some of the talk that has arisen about AA not being a worthwhile exercise and stating that pixel art that is smooth enough to be vector art shouldn't have been pixeled at all just did not sit quite well with me.

Perhaps that's because it didn't actually occur?

Helm's statement

it can be said that a tour de force of super-AA smooth pixel art seems to devolve: it looks like vector art and one has to wonder why it isn't vector art, then.

.. can you honestly say this truly expresses the sentiment that 'pixel art that is smooth enough to be vector art shouldn't have been pixeled at all'?

It's kind of like dithering is a type anti-alias arranged in a tight repeating pattern, if that makes sense.

It's been observed here on Pixelation in the past, that dither is the literal complement of AA : AA trades color resolution for apparent spatial resolution, dither trades spatial resolution for apparent color resolution. I wasn't sure what to make of your post, but perhaps you'll get something out of this idea.

EDIT: most of the most lauded pieces which are heavy on dithering are in category C, btw..

[Dex]

Perhaps it was just the language used, then. I don't mean to start any arguments, of course-- just trying to voice my thoughts to make sure that the inherent ideas of pixel art aren't being challenged.

Over time my pixeling style has worked to attempt to eliminate banding in any shape or form, and I attribute that to the studies done both here on Pixelation and how I've learned and grown over time as well. I know that the idea proposed here isn't to truly eliminate working with single pixels, but some of the talk that has arisen about AA not being a worthwhile exercise and stating that pixel art that is smooth enough to be vector art shouldn't have been pixeled at all just did not sit quite well with me.

Perhaps that's because it didn't actually occur?

Helm's statement

it can be said that a tour de force of super-AA smooth pixel art seems to devolve: it looks like vector art and one has to wonder why it isn't vector art, then.

.. can you honestly say this truly expresses the sentiment that 'pixel art that is smooth enough to be vector art shouldn't have been pixeled at all'?

This does, however, still seem like it's splitting hairs about how pixel art should be used, though. I understand that it is sort of a "what if" question and not adamantly trying to discourage incredibly smooth pixel art, but I've seen similar posts thrown around about demoscene pixel art or other incredibly detailed pixel works (whether they be close to realism, digital paintings, vectors, etc.) that often consist of "this looks like X kind of art. why didn't you just do it with X?"

Should we question how and what ways pixels are utilized? I think that is an interesting question, honestly. I mean, we do of course to an extent. We don't qualify oekaki-type art as pixel art, as the fundamentals of pixel art are "control over the pixels" so, yes, we do question in a sense how pixels are utilized. How should we apply this question, though, and to what kind of pixel art? And if so, is it necessary to do so?

[Ai]

This does, however, still seem like it's splitting hairs about how pixel art should be used, though. I understand that it is sort of a "what if" question and not adamantly trying to discourage incredibly smooth pixel art, but I've seen similar posts thrown around about demoscene pixel art or other incredibly detailed pixel works (whether they be close to realism, digital paintings, vectors, etc.) that often consist of "this looks like X kind of art. why didn't you just do it with X?"

Assuming that you're saying it out of curiosity and not smartassitude, that's a worthwhile question. Why didn't they? Is it because they get something personally out of that particular method? Because it helps them structure the image better? Because it's actually harder to achieve by the other means? Because they're deep in their comfort zone and are tempted to avoid new methods?
All of these things have implications for their development as an artist.

I feel 100% confident in saying: Helm is very not* a prescriptivist and his question is indeed posed in the spirit of 'what if', much as a majority of his posts here are. He's a careful thinker.


* grammar is correct. or at least intentional

Should we question how and what ways pixels are utilized? I think that is an interesting question, honestly. I mean, we do of course to an extent. We don't qualify oekaki-type art as pixel art, as the fundamentals of pixel art are "control over the pixels" so, yes, we do question in a sense how pixels are utilized. How should we apply this question, though, and to what kind of pixel art? And if so, is it necessary to do so?

I think that it's necessary for the artist to do so -- for their own art, be it pixelled or not. All methods are simply proxies for our actual intent, so if we repeatedly use methods that are contrary to our intent, that tends to become a problem of underachievement / self-sabotage. No methods are exempt on any level from examination, although methods that are currently effective have an indefinitely-long stay of execution

For other artists, it's also necessary as a mental exercise, IMO (understanding others' art in order to expand their own repetoire).

It can also be used as a simple 'who's (not) in the in-group?' exercise, and of course I'm as against that as any other form of tribalism. But we can't really take the literal text of questions as an indicator of whether the question is honest or merely political. Context is necessary, and I believe in-context, Helm's and Cyangmou's questions can be seen as clearly in the general spirit of an inquiring mind.

A last thought:
You could view the idea of developing the ideal of X (pixel art, cubist art, digital art, sculpture) as basically about viewing the medium as a method for teaching yourself certain essential aspects of art. In this sense it's important to define it, in order to determine what exactly that learning curriculum will consist of. You wouldn't want to spend 3x as long as you needed on AA if the subject really consists largely of learning how to tightly+cleanly fit shapes together in pleasing proportion, for example. So this is the case for discussing definitions as a community: because it will help us learn more effectively.

[Dex]

Complete and utter agreement with everything you've said in that post. I definitely agree that the question at hand seems more in the spirit of inquiry than anything else, but hey, that is a point I thought that should be brought up, lest anyone get confused with the intent. Helm and Cyangmou are both very respectable and their questions are surely just ponderings, but I thought I would bring up the other side of the spectrum and question the... questioning in a sense?

I also agree with your perspective on the intent of asking "why was this created with this medium when it could have just as easily been created with this other medium." I believe I've seen both cases, honestly (asking out of curiosity, or asking out of "smartassitude") and thus I suppose it is entirely up to context. I think we should be careful, then, how we go about asking that particular question.

But yeah, you've essentially answered and addressed everything I've said quite well. I guess my next real question at hand is this: does this exercise in working with eliminating single pixels seem to be something that should be incorporated more often and on a regular basis, or is it just truly an exercise-- an excursion into a different way of thinking about the pixels?

[Helm]

Dex, I understand your concerns.

Here's the deal: Pixelation has in the past prescribed what pixel art should be. It could be argued it was necessary at that stage of the craft's maturation process - there was the introduction of the FAQ, and a further amendment of the FAQ. I was leading the charge on that stuff.

It wouldn't surprise anybody here to learn that I have issues with control. Nobody ends up administrating a community like this without being a bit of a control freak.

However, I have also grown as an individual and I've let go for a lot of what I considered worth controlling in the past. I completely changed the way I dealt with problematic individuals in Pixelation, for example, trying to have a conversation with them (usually behind the scenes) to address their grievances.

Likewise, there was a perspective shift when all the fuss was kicked up about 'NPA' (nonpixel art) over at Pixeljoint. I could tell that this was a feedback loop from earlier Pixelation days and I could tell we should break out of that and for the most part we did.

The way we did is by not prescribing what pixel art ought to be, just to recommend what people bring to Pixelation because that's what we could help with. This might seem like a fine distinction, but it really wasn't and isn't. It's the reason low-spec was also introduced.

The complicated issue, and the one Dex is correct in underlining occurs when I started getting into the cluster stuff, because although I was and am very clear that it all is just personal thoughts on the matter, the way I understand that stuff still has an air of 'this is what pixels long to be'. So to be extra clear for all intents and purposes, it is philosophically suspect and should be taken with a grain of salt, that 'pixels long to be' anything.

People are free to do oekaki, to have banding in their art, to have their dithering touch their aa, to do pictures with 3000 colors that took 500 workhours to finish. I am fine with all of that. We might not have something valuable to say about all of these things in Pixelation, but that shouldn't stop them. If anyone patronises them in Pixelation about it, let me know and I will make certain that it never happens again. It's one thing to be told 'I can't help with this' and another to be told 'you shouldn't do this because I can't help with this'.

So, what we discuss in this thread, it's best to take it as a vice. Those that have the kink will get what we're discussing and it will inform their art when they try it and they will not be able to help themselves but think of the repercussions of the suggestion of no single pixels. Others will look at this, go 'whuh?' and never give it a second thought, and that's absolutely fine.

Because we can still talk about what's going on in their art if there's single pixels, aa and dithering. We have the tools, we built them systematically for years. There's no reason to push people to adopt a kink they don't enjoy already.

So it's very different from oekaki (where the discussion usually stops at 'you should control more') or supersmooth demoscene artwork (where the discussion usually stops at 'wow! I can't believe it's made with pixels!!!!!!!!') because we can still talk about single pixels, see?

There's a reason I haven't even included this recent thread in the Ramblethread proper. It's because it's a vice, honestly. It might be an important thing in my mind (as all vices are to those who have them) but it's not necessary for good pixel art. At least I don't think so. So it's prescriptive, but only prescriptive to those that come to this thread and their brain goes "...of course." It might not go that way immediately, but you'd have to be game at some point anyway to entertain something so ludicrous as taking off all the single pixels from a medium built out of the love of the single pixel.

Rest assured, I'm fully aware of how crazy some of this sounds and I'm not going to enforce it in any official way. In fact, I'd say that I'm on the way out of Pixelation administration anyway. It's been too long and I feel I'm holding the place back. I'm ready to be the crazy uncle, mumbling about clusters in his sleep by the fireplace.

[API-Beast]

This thread makes me realize that how much obsessive anti aliasing can hurt the piece since it forces you to pack a lot colors into a extremely small area.

[Cyangmou]

Dex I can completely understand you.

It's not up to anyone to judge "that's good art, that's bad art" - everyone has his own taste and his own criterias to judge about certain pieces.
And I think that the whole judging thematic destroys artistic freedom on a very basic level.
If we start to judge something "as community", and even if it's just voting, it's an harsh insection of an "individual" artists freedom.

The only thing we can judge in art and we can measure is realism. The more realistic something gets the more we can compare it with reality and the easier it is to judge about an artists eye (Gottfried Helnwein?).
"oh wow that's like a photo" (craftmanship?)

However if we talk about something abstract and we don't get the ideas of the creator behind a piece (Wassily Kandinsky?)
 there is a chance that we don't understand it and see it as "trash" without even questioning what's behind it (ideas?)


It's all up to oneselfs ideals.
Narrow-minded people are fearing the unknown. A lot of artists we consider as great nowadays brought something new to the table other artists don't understood and it was easier to say it's bullshit than trying to reconstruct a certain thinking process.
The problem is that some great artists aren't even questioning their perspective and they are doing stuff because they learned it that way and always will do it that way.


Now the big overall contradiction with pixel art technique is that it nowhere has anything to do with common "artistic craftmanship". It's not like perspective, color theory or anatomy and all of those techniques we need to make a descriptive piece of art.
We can't even measure it against anything
I suppose it's a much more abstract way of thinking about how to create a piece with pixels and how we use those pixels to get an effect.
And we need experimental pieces to look at to get out something.



Elks dragon would look great with another technique as well, same goes for Panda's falcon. Is it great "pixel art" (with applied pixel techniques we discuss here in this thread) or is it great art made with pixels (artistic craftmanship in the common sense, like crowd-voted on Pixeljoint)

Furthermore and I don't mean this insulting in any way (nor would I have brought it up, but it serves as a great example for "perception" and we are particularly talking about that in this thread as well and it lines up perfectly with the first paragraph)
 
your Secret Santa piece.
http://wayofthepixel.net/upload/ss13/tim_from_Dex.png

If I measure it with my own preferences I'd say the effort you put into it shows off and it's something everyone could get inspired from, the colors are great, the subjects are well chosen and the overall look is also nice.

In terms of common artistic technique the piece fails in nearly all spots where it wants to shine.
the most obvious wrong spots:
In terms of anatomical details are both elbows,
in terms of posture it's how you connected the lower torso with the legs,
in terms of proportion it's the neck,
in terms of drapery it's the piece wrapped around the upper arm
the animal anatomy of the pigeons is quite off to, but that's a side detail

If we will crowd-judge this piece at Pixeljoint It will be percepted well, because a lot of people just see the overall details but can't look through the basis of the image or they lack the in-depth knowledge.
Nonetheless the piece definitely will find its audience as every piece with a certain level of skill does.

But just because the overall perception will be well this won't mean that I will change my opinion about it or I will adjust my knowledge to the average one.
And I could be nice and say "oh wow it's so great how much effort you put into it, how well the colors work and how great the overall impression is" then I also won't lie, I'd just have shifted my usual priorities

Whenever we judge something there is a serious danger that the creator adjusts his own artistic vision to the judges opinion which takes intentionally or unintentionally away from artistic freedom.




In arts we first learn about certain rules (or at least we should gather enough basic knowledge)
and then everyone goes his way and breaks them intentionally to his preferences
Helm was questioning the basic ideas and ideals of pixel art and I think everyone can gain from it, if we are capable of working it out.

The technique (or ruleset) we are discussing here has not a lot to do with our own judgment criterias and we can't apply it 100% to our established techniques yet.
It's an experience to apply different criterias to a piece and to look at it from a completely different perspective.
And from experimenting you can gain experiences and with experiences you can grow.
With my provocant questions  like "leaving out AA completely" I just want to emphasize that everyone who participates in this thread overthinks his ideals as some people already did.

After all my experiences so far, I'd heavily recommend that you don't apply this technique to your old art first,
instead make something completely new which don't necessarily has to be "a part of you" and it should get a lot easier to experiment with a different mindset.


@r4c7:

Yeah you exactly brought up what I stated earlier - for me the key element also is the contrast.

[Arne]

I think these kind of "rules" (practicalities?) are nice to have in the debugging/refinement tool belt. Cluster control/noise reduction is definitely on the Top 5 list of things I check for when a piece isn't working for me (in particular because I do so much lowrez game art). I think it's an expression of a similar idea that I encounter when doing concept art (don't do tons of stupid little random details which break up the shapes and obscures the punchline).

I do think pixels long to be something, but in a different sense. It's so easy (for me, and I think others) to fall under the spell of the pixels and let their very nature dictate... certain structural embroideries which does a piece a disservice. I have to yank myself out of it constantly and consider seemingly crazy ideas like asymmetries, omission, illogical relocation, absurd adjacency...

[ErekT]

Okay, tried retouching another old image. Here are some observations. Not using any single pixels forces avoidance of some mechanical patterns. The most obvious example being dithering. But also certain AA-solutions that I tend to fall back on a lot. It makes for more organic solutions.

Things do get smeary though, and you lose some of that minute-level detail that's also a big strength of pixel art. Feels like it's not something that works well in 100% of cases, at least for me. But I like it, it makes me re-think several not so good habits I wasn't aware I had.

[Helm]

I'm glad the concept is helping your art. If you do this for a while, then later if you allow yourself a few single pixels they'll feel even MORE sharp than they do usually. The idea is to internalize this enough so you don't always reach for the single pixels.

[RAV]

I weigh in once more, because this time it might be more directly relevant to you.

As you should understand by now, my voxel tool revolves around truly Composite Resolution, as integrative workflow on-the-fly.

And as I have layed out before by the example of 45°, this can create for a duality of understanding, that yet is cooperative.

The same is true for "single pixel". When looking at it from the finest resolution point of view, this term names a problem here, because it encourages certain techniques that maybe should not be the first thought upon setting out a work, but the last if anything. However when you consider dynamic resolution switches as part of how you create art, you find that a single pixel of a lower resolution is a cluster of its own on a higher resolution. And so you add single pixels on each iterative layer of resolution, to sculpt the pixels of the last, until on the finest you end up with a very cluster oriented construction; although this dynamic resolution workflow incorporates "single pixel" action, methods like dithering and AA become an afterthought at the end if anything, and even what you call 45°, because all of this turns out unnatural and more work in this kind of workflow; so maybe you could even say this innately informs the artist on pixel art as described here, without really enforcing it.

In his latest piece on PJ, Drazelic tried to emulate this workflow somewhat, though it is less intuitive without dedicated support:

by doubling-size and then re-pixeling over the previous version to bring it back to 1:1 pixel resolution.

I don't know if this particular result has much to do with what you are after or examplifies my point here, but at any case I think its interesting that we have these two unrelated techniques developed, that in their interpretation and terminology even suggest opposition, and yet maybe end up being a good fit.

[ptoing]

I did not read the whole thread, but some very interesting stuff in here. I wonder how much different working with clean clusters is when not using 1:1 aspect pixels, like C64 MCOL stuff with this kind of approach.

[Cure]

Am I doing it right?

[Helm]

Yes, how do you feel about the result?


RAV, do you mind showing pixel art that demonstrates your technique?

[RAV]

Ah, my friend, you know I'm not an artist. I thought that Drazelic's animation gave a good example of what I tried to explain. But I try something.

I don't know what's right or wrong here, just that my curiosity is honest, and my intuition tells me that although we started from different points of thought and process, both our things point into an eerily similar direction; take it as another confirmation from my unlikely side that this "direction" explored here might really be more than just a crazy quirk for nerd-iest of artists.

So now I will embarrass myself for your service: On this board here I am probably the most unskilled artist by far. The following is pretty much the first time in my life I "draw" something that might qualify as pixelart; or that I draw anything; last time I did was an oekaki of a *shoe* for when joking around with Cure -- a year ago. And before that nothing for a decade maybe?

If this here is supposed to teach people about pixelart, then I am the perfect guinea pig. In my success or mistakes, you can study on me.

So this is really quick here, I am "abusing" my voxel tool for creating pixel art the way I tried to describe. Sorry in advance for wasting five minutes of your life:



Now I want you to realize what I'm trying to say: this is not any "groundbreaking" process; but what I'm trying to hint on is that I believe there is quite a difference for reinforcing the learning process of what this is about here, between just blocking out with a different brush-size on finest resolution, or blocking out in a native lower resolution iteratively. I think that the later has an even greater effect on informing the principles of conscious polyomino-fication, from macro to micro, like a fractal.


You can observe two things readily: I start out with single pixels and 45°, but also on the next step, I already made my first few polyominoes; what I didn't show in the end was this: For making "texture" like little bumps, cracks and ornaments, I would have gone high and higher res, and distributed single pixels on blade and shaft, before I would go yet even higher res step by step and mold those pixels into polyminoes of themselves, etc. There is even that nasty orange 45° on the shaft at one point, but this one too would have "dissolved" in "resolution pixel sculpting". Same for the "backround".

I don't know if this helps by any means, maybe what I say is for granted already, although what I'm trying to say here is a little distinction, a little spin on this, that I think reinforces this concept as a whole. Or maybe what I'm trying to say is awkward bullshit, but by that it might help you too proving or better explaining yourself? I'm just really curious of how this turns out because I deem it relevant to my work, so I try to help, and maybe better understand this on my own too, although I'm no artist, but how to make good tools for art, if I have no clue about art, you understand my motive here.

[Cure]

Yes, how do you feel the result?

On some areas it has cleaned up excessive aa and I enjoy the simplicity, overall it feels like trying to pixel with one hand tied behind my back. It certainly has a more 'pixelly' look, it's a bit low-fi for me though. It's satisfying to make clusters click, so it's probably a worthwhile exercise on less complicated images.

I will say that while it suffers a loss in fidelity, the actual pixel relationships are much more interesting.

[Ai]

RAV: Now that I understand what it is you are trying to say, I definitely agree. I have tried that iterative scale refinement process before.
Your video inspired me to try it again.

These are the stages 4, 2, 1. (that is, 4x4 pixels, 2x2 pixels, 1x1 pixels). I think I started at 8x8 pixels. It should conform mostly also to the stricter (90degree) version of the cluster exercise.

It really felt like the art was doing me, rather than vice versa. Disturbing. Well, it's probably just 'ego-death'.

Anyhow, my assessment of it is:
* does force more control
* I feel more anchored to the existing things in the picture, less willing to change them.
* more consistent positive feedback loop
* I struggle to avoid making things look sticky (look at the image and you may see what I mean)


Howto for GraFX2:
* Set your picture size to whatever you intend the final res to be.
* Set the grid size according to the pixelsize you want to start with. Turn on 'Show' and turn off 'Snap'
* Draw a non-bgcolor rectangle that exactly fills a grid cell.
* Grab that as your brush, and rightclick on the brush icon to set it to monochrome mode.
* Click on the lower 'fx' icon and set the brush handle to the upper left.
* Open the Grid dialog again and turn on Snap.
* Pick your color and start painting.
* When you want to go to the next finest level:
 ** Adjust the Grid size accordingly
 ** Hit 'h' to halve your brush size
 ** Set the brush handle to the upper left again (for some reason, 'h' resets the brush handle.)

4-step version:
* set your grid according to the current resolution you want
* have a brush that matches your cell size, set to monochrome mode
* make sure the brush handle is not central (I recommend top-left)
* draw

[Helm]

Yes, that video certainly helped illustrate your previous posts. Thank you for making it.

It's certainly the most literal way to work from general to specific in the pixel art medium, and I theorize that it is a useful activity - I also am dead certain that it will create exactly the problem Ai experienced above, where you're not likely to fundamentaly change an image as you go because you've already gone through a few sculpting zoom in iterations and psychologically you might be more tied to it.

However that is a larger scale issue with pixel art, and especially if there's no single pixels and the cluster relationships are super tight, small changes have a ripple effect. Large changes... you might as well delete a whole part of the image instead.

Like... the first thing I usually do when I edit people to help with their stuff is delete all highlights on their sprites, go back to middle color and fix from there (it's no wonder people have trouble with light - it's much more complex than a simple shadow cast). If I try to edit large highlight clusters while they're already there, I do a bad job. This is a thing you only learn from editing other people's work - when I work I don't even get so far lost in a bad highlight that I have to roll it back - but I do spot, say, bad anatomy waaaay down the line (not enough anatomy practice   ) and with this no-single-pixels deal, I'd have to delete large parts of the image to fix that.

But so be it, this is something we have to get over as pixel artists anyway, might as well learn early.

[Helm]

And a further thought on the topic.

Back in the ramblethread preamble (don't make the obvious pun) I wrote that one of the purposes of pixel art is to convey a higher resolution than what a single pixel really is. I don't agree with myself anymore, not even insofar that there can be said to be any 'purpose of pixel art'. Even that aside, I tend to lean more towards making the intended viewer resolution *slightly larger than 1=1 ratio* and then using the 1=1 pixel as a punctuation mark.

If we consider a single point of AA to attempt to convey a subpixel effect, meaning, a half pixel, here and there, then we have an image that is *mostly* 1 = 1 ratio of resolution, with gimmicks here and there that are 1 = 0.5 ratio. This is basically mixed resolution. Of course it's not as bad as actually mixing resolutions, it's mostly ok when looking at it from afar and if the picture is lush in colors anyway (lower contrast in the touching clusters, as well observed in this thread) but if you zoom in, it starts to look kind of like patchwork. Imagine someone making lush oil brush strokes for most of his art, and then going in with a flat marker and puncturing points here and there.

I theorize that even subconsciously this effect is noticed at 1x zoom, though I don't expect anyone to come up to me and remark on it. This is the gimmicky quality that I now see, but it doesn't have a name and it doesn't exist in the culture. Perhaps it's just me, also.

If we consider, then, an image made mostly out of at least two-pixel clusters, with the very occasional, super-sharp single pixel effect, there is no gimmick, and the effect of sharpness is retained and used wherever needed. No loss of coherency happens on zoom, there is no effect of 'mixed resolutions', as a single pixel lives just fine in the tetris jumble if needed be, and it doesn't have to be there either.

We live in 2013 (in 24 hours 2014). Most screens that will show pixel art today are crisp, well-lit, no bleed, high resolution screens. Do we really need, for future projects, to 'fake' less than 1 = 1 pixel resolution? And then we might even blow up the art to 2x or 3x zoom to show people how clever our aa and dithering is.

I'm not saying it's 'wrong'. I do it for work every day. But I am saying we should be thinking about it more.

Cure says above that there is a loss of fidelity. He worked over an image that is a cleaned up oil painting, so his eyes have already seen a level of fidelity pixel art would not be able to achieve in any case for this image. I think that at 1x zoom looking at them flip between themselves, I prefer the cleaned up version. It's sharper and much better understood without zooming and most importantly strong lines make much better visual sense as 'staircases' than as banding-by-themselves-lines. I theorize for pieces made from the ground up in this mindspace, without any mind's eye higher fidelity to sacrifice, the artist will not struggle to convey something with two pixel clusters as much as they might think (granted, with a bit of training).

And all of this is not to say the two-pixel cluster is the new pixel. Even those clusters shouldn't be overused (I'd have to practice a lot ot see what that means intuitively for my own art). There should be a healthy mix from broader clusters towards the smallest ones. Again, brush stroke economy from real fine art training would be useful here.

[PixelPiledriver]

Interesting thread.
Just wanted to throw in some reading that I was reminded of by your last post Helm.
http://msdn.microsoft.com/en-us/library/windows/desktop/cc627092(v=vs.85).aspx
It may or may not help the conversation.

Once you start to involve numbers into anything its good to do quantity tests.
Raise the cluster size minimum by 1 and keep going.

I think I've thought about this before and forgot to do it:
It would be cool to program a line tool that has parameters for min and max length.

It can be visually repetitive, altho only subtly so, to add a single aa pixel to edge of all faces.
But I think that's why it's easier to perceive and apply aa as edges of vectors that fall on a pixel map.
That way you get thinking more about creating indentations, bulges, curves, weight etc.
Also AA feels a lot different in motion.



Edit:
Here's one more with a few changes to make the last part dissipate better.

[RAV]

Yes, this thread is what I deem the most worthy way of concluding this year and starting out the next, thanks everyone participating.

Helm, hopefully without hijacking this thread too much, I want to give some more explanation of my thought process, why I really did what I did, although it's not immediately relevant to you now, but I think it's interesting, because in a way it's like reliving the early start-up progress of computer graphics / pixel art all over again; that's what makes this "new frontier" kind of exciting to me, it's all the same limitations and problems and aesthetics in the next dimension.

In what is populary referred to as voxel art in the form of cube meshes, resolution is as much a problem as it was for pixels on early computers back in the day. A higher resolution exponentially increases processing and memory requirements in pixel art, and for voxels it's sooo much worse even, at power3, making it extremely demanding even for today's best computers.

So what people do in things like Minecraft is making it extremely low-fy, and trying to make it more readable with higher res textures flat on them. In a way this is the worst kind of mix-and-match resolution. And then there is the idea of, hey why not give different classes of items a different resolution uniform, like the terrain is really rough, and the characters a bit less rough. This is a bit better, but still quite some hodge-podge.

So what I do is an actual natively mixed resolution on the fly for everything. BUT the trick is, the ace card, the joke of it, that it is not really meant to encourage mix and match resolution on the resulting art work! In fact what I really intended it to be used like is to give all the world a uniform fine resolution unlike any other such project to date.

What you do when modeling cubes in Blackbox is *surfacing* it  ;p. The layers of lower resolutions turn out to be hidden within the core of the model, or away from sight on a walled side not meant to be seen, like a facade in film and theater. They are just a logical requirement for constructing the model, or modifying it later, but they are not meant to define visible model aesthetics, rather as a purely logical definition of volume, and to save so much performance and memory that it makes a higher surface resolution and frame-based animation and large distances possible on today's computers.

They achieve this through dynamic level of detail, that means the engine itself decides in a given scene, how much detail of each object is shown. When you put your face down on the ground tile, you see every stick and stone in ten thousand fine cubes of surface "texture", but when you look at that tile from a mile away it's only that small hand full big cubes of its core, but you don't really mind/notice, because it's so far away, the perspective makes it so small as a couple real pixels. As you get closer, it fades back in layer on layer of cube-resolution. Important in my implementation is, that the engine understands that the bigger cubes can at any time be transformed/split into smaller, to make all assets dynamically modifiable/destroyable, as if it was all a single uniform resolution at highest; that means that resolution-information is not actually lost, it just knows how to abstract things for the right occasions, to get the best of both worlds.

In normal pixel art there is no "hidden pixels" as such; but if computers were less powerful than today, and we still wanted to do Doom-like sprite-based 3d, then the equivalent to my technique would be what's called mip-mapping, that means creating each sprite in multiple pixel resolutions for the engine to pick as required. And even though the construction of pixelart does not depend on hidden-low-res pixel either, it is really interesting that the technique we currently discuss actually does rely on exactly that... the mip-maps would be a natural by-product of how you create art anyway. So if this were the technical reality, it would naturally be the preferred way of creating pixelart.

So that's my little story, and the interesting historical point in time we're at, where in case of this technology, all our super modern hardware just turned into a C64 again, in terms of what is ever so required and what meager it can deliver, and how we have to work around all that by being smart, coder and artist hand in hand, rather than brute processing force of how any pixel art is delivered today as an afterthought by anyone who so much as sneezes. By that notion we might even say, voxelart of today is spiritually closer to pixelart back in the day, than modern pixel art itself, which, as has been noted before, in its lush fidelity is more and more difficult to differentiate from art made of pixels. But I don't mean that as valuation, I love it all, I mean that more in the modes of thought process of an artist in the process of creating. You know how you often put yourself under artificial restrictions to relive it and spark your creativity in certain ways? Well, these limitations are actual reality again for the next 10-20 years... maybe the last time any limitation is not artificial.

In entering discussion here, I was interested both in your advice, I am ever so doubtful of my devices, and if what I do has any value for you too, and if there are any funny quirks and interesting abuses and "syn-aesthetic" styles in it for pixelart as of now. It is always good to increase the total wealth of what pixel art can mean, in its pride.

[RAV]

Ah, another thing I should have mentioned.

When I talked about texture mip-mapping, and that what I do is basically geometric mip-mapping, I explained it purely from a performance point of view, as if making a Doom clone today would not make it necessary for the sprite art anymore. But the truth is, mip-mapping is as important today as it has ever been, it still saves a lot of performance in modern engines; their lower mip-maps are usually auto-generated by a tool, from the finest texture the artist provides; disregarding how the algorithm mangles pixels on downsize.

But the even more important reason is another, aesthetically: having a lot of textures on screen, all at the same finest resolution created natively, but rendered at different scales and angles by perspective, resized every frame on the fly as view and movement requires, can cause a lot of very nasty noise/interference on the screen, especially on many small and far objects that still render down-sized the largest texture version that normally would exceed the screen-space given by the object currently, it's a dirty pain to watch. But having a couple of prefab texture resolutions that fit more naturally to the different scales of an object on screen, and then just resize for soft transitions in between, greatly reduces that visual noise, and mangles less the pixels of manually created mip-maps.

For this reason, even if you made a Doom clone today, you would still manually create mip-maps for all sprites/textures, as natural by-product of the workflow we talked about here.
Also still important even in my engine, for creating a torrent of particle effects, with which volume doesn't matter, as pixel-sprites still.

And this is the same reason why I, other than for ever much more critical performance, do the level of detail for voxel-sprites, and require the artist to create it with that in mind, naturally supported by the tool, even if I had all the performance in the world, because attempting to render a lot of items in all their detailed close-up glory, even at far away, makes for just as much flickering headache for the viewer.

Frankly, even a classic 2d pixel art side-scrolling or isometric game, would still greatly benefit of it, if for some reason a lot of zooming in and out is necessary. Like a dynamic and fluent replacement for the classic "overworld" view of the whole map, or minimap.

Probably that too not directly relevant to you, but interesting to know maybe. Though I think it's enough of that for this thread, it should revolve more around the actual work with pixel clusters again...

[Lanarky]

I made Viewtiful Joe from scratch using DawnBringer's color palette, in IsoPix on my android tablet. I was trying to test myself with clusters and using a palette I couldn't change. ( I'll need to get used to touch screen. )

There are no single pixels in the image.

[Helm]

Very well done. How do you feel about the process and the end result?

[Lanarky]

I think the technique works really well, and I would suggest trying it first from scratch, since you have more control over the final outcome. There were a few places where single pixels would help, but it makes it easier to change areas of the same color with the fill tool. At first I thought this would only work with realism in pixel art, but with more practice I think it would work in more stylized approaches too.

I feel good about the outcome and I think I can see deeper into the structure of what I'm working on. I can spot single pixels easier, and banding pops out more when doing this. So I hope more people post some of their work since it seems like the thread is becoming a little off topic. I also started out thinking it wouldn't be good for anything but realism, but I think more people trying it out for themselves will make this a technique worthy of the ramble thread. You just need to step out of your comfort zone to be able to see it yourself.

[Helm]

It absolutely is not just for realistic stuff, this is something I did for work (I retouched it just now to have only two single pixels. Before it had 45degree connections)



It's a Bad Ice-Cream avatar for a game we finished up before the christmas break, and during this is where I realized the single pixel aa problem. I was trying to keep it sharp for 1x zoom which is mandatory for avatars and it dawned on me what I was doing with single pixels.

Btw you can play the game here http://www.nitrome.com/games/badicecream3/

I am working on something more ambitious as we speak.

Also you are correct, I'm finding that fill-tool changes make for a faster workflow than usual anyway. It's interesting

[Lanarky]

Another thing I noticed was, you can use a lighter color to do the "staircase" effect on diagonals to trick your mind into thinking its a single pixel outline at 1x zoom. Almost like a sub-pixel, where you have to think of making a bolder lighter line to make a thinner darker line in animation tweening.

[ErekT]

I wanted to try this on an image with an excessive amount of dither and single-pixel detail to see what happened. Kept a few single pixels for the neon-sign characters but the rest should be without unless I missed any. It's not as hard to preserve detail as I thought before I tried any of this.

[Vagrant]

I'd say it looks way better now Erek. 


Here's another from me. This time from scratch.



Original



Originally intended to be for the anatomy study thread. I used a picture Dennis posted about that Spiderman girl, and drew the skull over it. But then again, why not make it into clusters anyways? Two studies in one.

Maybe I'll finish it, just to have more stuff in my PJ account. =P

[Helm]

How do you feel about it now that you did something for the ground up for it?

[Vagrant]

Well, where do I begin?

Playing with clusters is a fun game in where nothing ever seems to go wrong, at least for me. It is similar to painting, only exchanging brush strokes for clusters, and some of the results can be beautiful even without single pixel AA here and there.

Hopefully the image will speak for itself, im burned out over here and too sleepy to type lengthy. But I'll come back to this piece and comment more soon. I might do the hands next time.

[Parkerbaby]

I love the game, Helm. It's unfortunate the main character is heavily banded.
I was trying to pick out the artwork you did.

Did you do any of the baddies' animations? They're great.

[Helm]

Yes, legacy art from two prior games I couldn't touch up.

I did most of the tilework for this one and I did all the stills for the new baddies, and animated the cow and the cactus block

[Vagrant]

I seem to be getting a grip on this technique; it comes now automatically the more I work on it.



There should be no single pixels in the skull now. Pure cluster goodness.

There are in the hand however.

Deviating into anatomy for a bit, do the shoulderblade/spine in the back look right? I've been doing from memory, not to mention finding a ref in that pose is not easy.

[ErekT]

Nice job

Single pixels in the hand and wrist look pretty fixable, I guess you didn't get around to it yet?

[Crow]

Nice job

Single pixels in the hand and wrist look pretty fixable, I guess you didn't get around to it yet?

The question is: is it necessary? To me, it doesn't seem to be the point to eliminate each and every single/isolated pixel that exists on the canvas, but rather avoid them where possible. That's no reason not to use them where it makes sense. Giving up on the smallest possible entity will make you lose quite some control over your piece, no?

[ErekT]

The question is: is it necessary? To me, it doesn't seem to be the point to eliminate each and every single/isolated pixel that exists on the canvas, but rather avoid them where possible.

Necessary? Personally I don't think so. I won't abandon the single pixel for future stuff I make because I still think it has its uses. But I'll be cutting down on using it quite a bit.

But as an exercise I thought the point was to push yourself to find new solutions to the regular one-pixel fallback whenever possible, even when you think a single pixel might be more ideal. And imo it's very possible here.

[Vagrant]

Yup, I was getting to it. But I have to agree with Crow.

The question is: is it necessary? To me, it doesn't seem to be the point to eliminate each and every single/isolated pixel that exists on the canvas, but rather avoid them where possible.

Necessary? Personally I don't think so. I won't abandon the single pixel for future stuff I make because I still think it has its uses. But I'll be cutting down on using it quite a bit.

But as an exercise I thought the point was to push yourself to find new solutions to the regular one-pixel fallback whenever possible, even when you think a single pixel might be more ideal. And imo it's very possible here.

It is possible, but it kinda looks like a blurring effect was applied. Single pixel AA effectively defines the structure loads better however, with those angles. It's simply sharper.

[ErekT]

Yeh, I might have gone a little overboard with the AA there. Here's another two.



All down to personal preference in the end what solution you like best of course.

[ptoing]

I have to say that the double pixel 45 degree lines and generally fattened up outline stuff does not look great to me in most cases. It works better in some, like the back of the head of Viewtiful Joe, than others, such as long 45 degree lines, imo.

Also the fact that a line like this should be lighter than the colour it is intended to be, due to rules of antialias and if not done right a diagonal fattened line will look significantly thicker than a straight line).

I like the general idea of keeping clusters efficient, but I do not see the problem with single pixel outlines at all (look at Mia's lovely work, among others), or with single pixel step AA. If you have the colours to fit in 3 pixels of different value for AA in a place which ideally should have 3 pixels worth of AA, then 1 pixel, 2 pixels, 4 pixels, or 3 pixels consisting of less colours will likely give a worse result.

I have not had the time to experiment with this more, so this is all just from analytical thought and not from proper testing, which I will get to once I am back home (mid Jan) and have some time for it.

[Mathias]

I couldn't find the difference.
There it is:

[RAV]

I have to say that the double pixel 45 degree lines and generally fattened up outline stuff does not look great to me in most cases. It works better in some, like the back of the head of Viewtiful Joe, than others, such as long 45 degree lines, imo.

Also the fact that a line like this should be lighter than the colour it is intended to be, due to rules of antialias and if not done right a diagonal fattened line will look significantly thicker than a straight line).

Beyond outline, for an example of how it can look pretty cool in general, see Mrmo Tarius's works. However you might call this already a style in itself rather than a style agnostic method.

I tend to agree though, which is why I made it an optional tool for surgical operations. I just threw it in here to see what happens, but overall it's probably much less relevant to pixel art than "cube art"; in the next dimension it's more a concern about surface smoothness than outline thickness, since it can look really shitty when all those corners of cubes are sticking out like spikes, partially overlapping on perspective in an annoying way pixels wouldn't, the visual interference from that on the move, when there's a lot of it for intricate "texture". So I'm trying to get away from that ugly notion of "retro romance" that's driving that scene, and more towards an interpretation in creative terms of simplicity, control and tidiness, that makes visual sense and rekindles it with pixel art more on a philosophical level, so to speak. But in the end, how well that really works out is yet to be determined in ongoing research.

[Vagrant]

It's already influencing my own works.



But I'll still keep my single pixels where I need them. Now they are so much fewer.
Going back and re-doing an old piece can be a pain. It's most effective when you do it from the ground up, again.



Conclusion: Awesome thread and exercise.

[Cyangmou]

It's already influencing my own works.

But I'll still keep my single pixels where I need them. Now they are so much fewer.
Going back and re-doing an old piece can be a pain. It's most effective when you do it from the ground up, again.

Conclusion: Awesome thread and exercise.

Indeed
I also got much more sensitised to single pixels within my new pieces since I tried this exercise

 

[Helm]

I'm glad you are finding a use for this

[Cure]

I have to say that the double pixel 45 degree lines and generally fattened up outline stuff does not look great to me in most cases.
...
I like the general idea of keeping clusters efficient, but I do not see the problem with single pixel outlines at all

I agree with this. the fat 45 degree lines tend to be the clunkiest issue I come across using this method. This is a good practice for learning to create ideal clusters, but following the 'no single pixels' method too strictly limits the amount of solutions to pixel-pushing problems. In my opinion, it's best to use this method in moderation and still allow for plenty of single pixel usage. There is a difference between 'stray' or 'trapped' single pixels, and single pixels that serve a necessary function.


I've continued applying the no-lonely-pixels rule to Grishkin:

[||||]

I think that piece is really working well 'clustrified' Cure.  Neat to see the two versions together.



Just out of my own curiosity, this isn't considered a cluster is it?


I would guess it's two single pixels but I'm not 100%  positive.

[Helm]

It is if you allow 45 degree connections, but it's a bit of a leap.

[||||]

It is if you allow 45 degree connections, but it's a bit of a leap.

 

Thanks for the answer.

It's a very difficult exercise. I'm trying to apply it as go right now and I find many reasons to keep single pixels; but it's an interesting concept and I like most of the results in this thread.

[Vagrant]

It's about time you jumped on board here ||||

 


Don't do diagonals. It's what I did in the first try and it was merely just pixel swapping with no real difference.
Try beginning one from scratch if you can also.

[Lanarky]

Yep diagonals are hard without single pixels. But starting from scratch really helps with understanding clusters.

[r4c7]

It is if you allow 45 degree connections, but it's a bit of a leap.

What about multi-color Clusters? I feel if the contrast is low enough it could work although, it would probably look better fixed. I whipped up an example with two 45 connections. In this case, they could be easily solved, but in a more chaotic situation, they may not be able to. Its a bad example, but I don't have anything better on hand.

[||||]

R4c7 not sure.. but probably not unless you're allowing those diagonals. I think to avoid it will be a challenge.

Vagrant, I'll start something from scratch; good idea. Now I just need an idea of WHAT to make, & give it a try.  (no diagonals)

Update:
I left some single pixels on purpose but only when I really felt like they just looked better; I will probably scan over it and mess with it more later with rested eyes:

For example a couple of the curves just seemed to need single pixels and two of the stars have an intended singles.. there are probably a few here and there elsewhere.

 I was so delirious last night when I posted.

UPDATED again..  No Singles! ('cept the two black outside of the oval outlining the hair)
I was really daunted at 1st but I'm really glad I tried this.


Dang it.. there are still a couple singles in the hair here and there.

[Helm]

r4c7 think about why you're pleading for various allowances for single pixels. Be brave 

[Lóng]

Gave this a try on a smaller scale to test out how limiting it is, it was a fun exercise.

[rikfuzz]

Massively not a fan of double-thick diagonals on outlines, ouch.

[Pix3M]

I don't have a lot of pixel paintings lying around but I have my internet face....



What did I learn from this....

I noticed that an art style will look more pulled together if AA is made up of clusters, just like everything else. Mixing single pixels and clusters is now gonna start to look more sloppy.

Interesting.... I'll see what happens if I start going cluster nazi in future pixel paintings. 

[Jeremy]

Applied the technique by completely starting over, it helped to cut down on that single-pixel noise the original had but I think it falls pretty far short in terms of AA and other subpixel stuff.

E; forgot to mention I cheated with those 4 pixels in the eyes

[ptoing]

I have been playing around with this a bit on some of my older images, nothing really to show, but I came to the following conclusions.

• It is pointless to do this in an image which has AA, because you totally kill diagonal AA and hamper the potential of AA in general, which in a lot of cases actually changes the shape of the curve you might wanna portrait.
• If you aim to make art without any AA, I still do not see how the no single pixels/no single pixel lines (esp this) helps your image.
• It largely seems to come down to personal aethetic preferences if you like this or not. I personally don't (the no-single-pixel at-all-and-weird-fat-outlines paradigm)
• Everyone should just try and pixel as clean or as messy as they want.

[DatMuffinMan]

Been lurking in this thread for a while, thought I'd join in. Here's an apple I did, no single pixels (I don't think so, anyway..) -

[r1k]

it seems to me that the exercise becomes less meaningful when you do it on drawing containing large open areas.  Like this apple, and even the cranium of the skull erekt did.  Just browsing the thread it feels like the excecise is more meaningful when you focus on small areas where we would typically resort to useing alot of AA, or theres alot of detail.

[Cyangmou]

now it gets interesting

where is the difference between that and Oekaki?
I'd personally consider this piece as Oekaki - what makes it pixelart for you?

[ptoing]

I wish people would stop calling stuff Oekaki. Oekaki just means "drawing a picture" pretty much and it relates to a bunch of online painting programs mainly done in java. You could pixel with those programs and people have, I have to some degree when I was bored. It is not really a helpful term. When does pixelart stop being pixelart and become "oekaki" or a "doodle" or "another kind of digital art"? In any case Oekaki is not a helpful term for this, unrefined pixel doodle is more appropriate, though less short and convenient perhaps, but also not misleading.

To look at the other end of the spectrum of what Oekaki can be go here http://oekyo.org/06/
Most of those are grid copied, some are not, however, many are pretty far from "doodles"
You can also watch animations of many of them if you click ★動画

[PixelPiledriver]

Just increase the minimum.

I wish people would stop calling stuff Oekaki.

I actually just went and looked this up as I've heard the term a bunch of times but never really understood what it meant.
And yah it's a weird thing to call a sketch.
A repurposed definition I suppose.

[DatMuffinMan]

oh no, that was not refined at all, just a sloppy outline. As for increasing the minimum, I assume you mean the minimum cluster size, which does work, yet loses a lot of readability texture-wise. I'd definitely agree with the meaningless-ness of managing clusters on larger pieces; it's neither challenging nor productive. I'll try to do the same on a smaller version of the apple later.

[PixelPiledriver]

I'd definitely agree with the meaningless-ness of managing clusters on larger pieces; it's neither challenging nor productive.

Clusters are edge flow.
Edge flow matters.
I see the point of setting a min and max, but is there really a point in setting a range limitation?

http://s3.amazonaws.com/estock/fspid10/13/08/16/9/portrait-painting-1308169-o.jpg
http://www.zeldauniverse.net/wp-content/uploads/2012/05/link-stained-glass-full.jpg
http://www.pixeljoint.com/pixelart/54911.htm

Here's some stuff I did with this thread in mind




Kinda reminds me of rythm tengoku.

[Probo]

what does the term edge flow meanexactly PixelPiledriver? ive not heard that before. anywhere i can read about concepts like that?

[ptoing]

Generally edge flow is a term used in 3D modelling when you describe how the edges/lines of a model flow to describe the surface.

Example image
http://tomparkersartdump.files.wordpress.com/2011/07/topology_breakdown.jpg

I think PPD just means the way that blotches of colour or lines/hatching in a drawing can describe the direction/curvature of the surface.

[Probo]

thanks for clearing that up ptoing!

This thread is a very interesting read, this is a technique ive never thought about. I do love AA and the single pixels myself though! Will you be sticking to this way of working from now on then helm?

[PixelPiledriver]

what does the term edge flow meanexactly PixelPiledriver?

Generally edge flow is a term used in 3D modelling when you describe how the edges/lines of a model flow to describe the surface.

Yes I'm trying to relate the ideas here to other ideas about art.
Probo, do some research about 3D art and why modelers talk about, think about, and practice edge flow.
I think you will find that there are interesting similarities to what is discussed on this forum.
Give it some thought and try to apply it to 2D in your own way.

Does a single pixel lose importance as clusters get bigger?
Sure.
Do the shape and arrangement of clusters lose their importance as they get bigger?
I don't think so.

This thread seems to be about making things fatter.
Thicker.
Increasing the minimum.
Size doesn't seem to be the issue.
If you scale the total image and maximum cluster size, then you can scale the minimum to match.

An image can be seen for what it represents or what it is made up of.
And perhaps as it gets larger it is easier to see the object drawn and harder to see the chunks of color that interact and bring it together.
But it's still there.
And seeing the chunks makes it easier to control.

[r4c7]

Sorry I haven't been able to follow up on what I've said before. I've since then realized what I was thinking about was stupid, but I still feel single pixels can be okay. Examples with single pixels on left and no single pixels on right;

The thing I feel that warrants the most use of single pixels would to remove the jagged staircase shapes created. You can see this in the top right. I really hate the way this looks and it seems others do too. I would write more, but ptoing basicaally summarized what I would say in this post:

I have been playing around with this a bit on some of my older images, nothing really to show, but I came to the following conclusions.

• It is pointless to do this in an image which has AA, because you totally kill diagonal AA and hamper the potential of AA in general, which in a lot of cases actually changes the shape of the curve you might wanna portrait.
• If you aim to make art without any AA, I still do not see how the no single pixels/no single pixel lines (esp this) helps your image.
• It largely seems to come down to personal aethetic preferences if you like this or not. I personally don't (the no-single-pixel at-all-and-weird-fat-outlines paradigm)
• Everyone should just try and pixel as clean or as messy as they want.

I also redid another small piece in this style again. Also got the color count down from 8 to 6 too.

[ptoing]

I have to make a slight addendum to my above quoted post. The black/white/pink bottles that PPD posted are cool, and they show how you can utilise this approach if you use it as the foundation for a certain style, what with fat outlines and all that. So it certainly is helpful depending on what you do, but I do not feel like it works as a blanket approach.

[Dex]

decided to try something from the ground up using this method, and it was actually a lot of fun. not completely finished, but you can get the basic idea

[Cure]

There should be 2 lonely pixels, in the pupil and on the tip of the nose. Regardless of its limitations, it's an interesting and satisfying exercise. A single pixel is like an unpaired electron, it wants to click into a cluster.

[Helm]

Lovely, Cure. I think it's quite sharper.



I'm working on this, it's on a few different layers and it's going to parallax scroll when I'm done with it and learn how to do it. It's difficult to finish this when I have to pixel for ~8 hours a day at work, but I will, soon enough.

[Crow]

Lovely, Cure. I think it's quite sharper.

Is it really? I'm seriously questioning this statement. It's different in places, but both versions are extremely sharp and crisp and pleasant to look at.

[Mr. Fahrenheit]

Looks nice helm. What I'm seeing is that the space part looks like it was something rendered on a machine that uses tall pixels. I feel like there has to be a way that is more inventive then just 2 pixel tall clusters then what would normally be single pixels.

A lot of the areas besides the sky seem to have a banding, especially the ground, it hurts to look at!  .

Anyways, I cant wait for the developer of this theory to demonstrate exactly what he means.

[Helm]

The stars are double-tall because the starfield will scroll horisontally vertically at the end (it's a trippy effect). So they're brush-strokes that cohere to the direction of the movement. The ground isn't finished, of course.

Crow: to me, at 1x zoom one is immediately sharper than the other, but then again I've trained my eyes on this concept for some time now.

[ptoing]

[...] scroll horisontally [...] brush-strokes that cohere to the direction of the movement

vertically then?

Crow: to me, at 1x zoom one is immediately sharper than the other, but then again I've trained my eyes on this concept for some time now.

To me at 1x zoom the new one looks less consistent because it has parts which have AA and then parts which are very blocky because of lack of AA, for example around parts of the hairline.

Also, how do you define sharp? Is sharp in this case jagged edges, where I would agree. Because I do not quite see what you mean really.

Mr. Fahrenheit also raises a good point. On a system which has a sufficiently tall or wide pixelratio (or supports modes that do) there are no percieved single-single pixels.

Personally I feel that this whole discussion is extremely esoterical and almost entirely outside the realm of objective reasoning.

That said, that piece you are working on is great, though I still feel that the embargo on single pixels hurts more than it helps in certain places, especially when it comes to issues of AA. As I stated before, double pixel AA forces a different, often suboptimal, outcome when pixelling curves. Also your AA is kinda blurry in places. Another small crit: Her castshadow should not widen, but be either relatively straight or taper like \ / a bit. Probably more toward almost straight tho. Because for the shadow to be like it is the lightsource which is behind her would have to be pretty close.

[Helm]

Honestly, I do not see where objective reasoning enters the conversation over aesthetics.

[ptoing]

What I meant with the objectivity part is that if you look at other aspects of pixel art you can make objective statements. Like when using AA one should keep the "higher resolution" of an object in mind, as not to distort the original shape/curve of the thing you AA, but to aid it.

This is where I think the 2 pixel thing often does not help, because again, it creates, imo, suboptimal AA.

I see this kinda practise helpful as far as building a style around it goes, but same goes for other. One thing I would give this approach (not necessarily the no single pixels approach at all, but a generally cluster focused one) is that it improves workflow.

[Helm]

"What I meant with the objectivity part is that if you look at other aspects of pixel art you can make objective statements."

I never have had any interest in making any objective statements about how AA should be. My intuitions and the framework I've explained over the years in the ramblethread come from a different place, even though the language I've used sometimes can be construed as trying for objectivity (because I talk like a robot). Let there be no misunderstanding, I state it plainly now (and as many times more as I have to in the future, I will), what I'm talking about is about aesthetics, and aesthetics are personal, third parties either see something in them or they don't (or they do further down the line). There needs to be no consensus and no objectivity talk.

I find the concept of 'objective statements in art' to be fascinating sometimes but mostly a fantasy in application and do not wish to feed that approach to art. I urge you to find what it is about 'suboptimal AA' as you call it you find aesthetically displeasing because I'm certain it's not about anything objective. I certainly don't find, for example, banding displeasing because of objective reasons, but because I find horisontal/vertical breaking to look ugly.

Insofar as we can discuss things, of course we present arguments that try to be structured and logical. I can say why I prefer a staircase of pixels instead of a 45 degree single pixel line now, and you can assess that position and agree or disagree, that's all good - but I do not want to talk about objectivity in any of that, we are all bringing in our personalities in making aesthetic choices. Finer and finer apparent resolution is not a set in stone goal of all pixel art.

If you like order, things in a structure, then you might be used in a very specific type of order. This approach here that deemphasizes more abstract cluster connections (because that's what single pixel connections are - more abstract, the mind has to fill in more missing data) might upset you because it's a different type of order. Have you tried it? Have you pixelled anything in this style from the ground up, to see if your mind readjusts? I certainly would find this whole thing preposterous if I read it and hadn't tried it first-hand.

By trying to find a place for this thing by saying that the Rhythm Tengoku glasses PPD drew are okay for it (but not other things?) you're trying to impose some order on something you don't have a lot of experience with. The picture I'm drawing has nothing to do with Rhythm Tengoku, and you might not like it and you might find things wrong with it, but they are choices I'm making on purpose to achieve a different thing than what you're expecting (which could be achieved with single pixels).

So, in all, I would suggest making a few pictures with this method before you write your opinion in ink. And try to be brave about them, they don't have to be ink-drawing iconography like PPD's glasses. I never got into pixel art to impose order on something, I got into it so I could cultivate some impulses in a form on which I had the most control.

re: critique on the image, thanks.

[ptoing]

I agree with most of what you said, very much actually. I have not done any of this from the ground up, because of being busy atm, tho as I said I tried to apply it to older works, and just felt that they got worse.

I did not mean to say that it is not good for anything else other than the kinda stuff PPD did, but that it would only really work if the whole image was done with it in mind from the ground up, as your woman in front of starfield picture shows (which I like a lot so far).

The thing about order is true to the extent that I have certain tendencies when it comes to neatness in some fields, such as consistency. Consistency is something that is important to me, maybe more than it should be. But from that point of view the new version of the image Cure made, to me looks less consistent and as such less aesthetic.

When I say "suboptimal AA", I mean AA that does not best suit the form that you are trying to show, and this could be backed up mathematically if you were making a curve for example. You will have cases where single pixel AA or AA made from a bunch of single pixels of various shades will yield a better result than forced double pixel AA. So I would say that, yes, there is something objective about this claim.

I guess the closest I have done to this kinda approach is probably my MSX stuff, which still has single pixels, but also a very focused approach to clusters in general, both because of necessity, but also preference of neat clusters.

[Helm]

But do you see why for other people it might not be important to be mathematically correct in conveying a curve, or at least as important as it is to not have to use single pixels to do it? Because they might be losing some other coherency in their picture by doing that, and they might find that to be an uneven sacrifice?

This is what this is basically about. I'd rather see fat 45 degree swords than skillfully subpixelled complex curves in 16x16 pixels. I can - and have in the past -been dazzled by people who can mathematically find solutions by hand on how to do this - I can even, sorta, kinda do proper AA if I try really hard - but the problem is - I suspect - that the facade falls apart when you zoom, the image loses coherency, the pixel islands become too far apart, information becomes stranded and my mind strains to connect things that I wanted to be connected on their own to begin with.

Mental challenge: try to think of a pretty vector piece. Zoom in on it gradually in your mind, every shape still being coherent even though everything becomes more abstract and less like 'a real thing' the more you zoom in. Distances between shapes become greater, but the shapes are still readable.

This is an element you would expect for there to exist in an art made of atomic bits, a pixel art. But it doesn't. Because we do not treat single pixels as clusters, but as tapers and connectors of bigger clusters - or even worse, buffers between two large colors (1point aa). We abuse them and they look like burning dots on a canvas the more you look closer.

I think the perfectly AAed curve is a big sacrifice to make to destabilize something in zoom-in. But if someone has that desire and that ambition, I'm not going to tell them to not AA.

This is a thread for people who, regardless of how crazy this might all sound in the beginning, after a little bit of testing, might have an 'aha!' moment. I'm not seeking to pressure people into converting into scientology or anything.

[ptoing]

But do you see why for other people it might not be important to be mathematically correct in conveying a curve, or at least as important as it is to not have to use single pixels to do it? Because they might be losing some other coherency in their picture by doing that, and they might find that to be an uneven sacrifice?

I can see that, yes.

Mental challenge: try to think of a pretty vector piece. Zoom in on it gradually in your mind, every shape still being coherent even though everything becomes more abstract and less like 'a real thing' the more you zoom in. Distances between shapes become greater, but the shapes are still readable.

I see what you mean there. Opposite to that you can also imagine zooming out, which is when you get stuff like AA being used as tapers and so on.

This is an element you would expect for there to exist in an art made of atomic bits, a pixel art. But it doesn't. Because we do not treat single pixels as clusters, but as tapers and connectors of bigger clusters - or even worse, buffers between two large colors (1point aa). We abuse them and they look like burning dots on a canvas the more you look closer.

Saying that we abuse them is a bit much I think, seeing as some people enjoy/prefer that kinda look. It all depends what you want to do and how you feel about it aesthetically, and it also HEAVILY depends on your target resolution. Arguably pixel art made outside of games or other practical application does not necessarily have a target resolution.

This is a thread for people who, regardless of how crazy this might all sound in the beginning, after a little bit of testing, might have an 'aha!' moment. I'm not seeking to pressure people into converting into scientology or anything.

I can also see this and I am sure I will have an aha moment at some point if I sit down and do some ground up work in this approach. Clusterology sounds funny.

[Cure]

tried one from scratch
the stars suck but everything else turned out ok.

[Faceless]

This thread is interesting to me, and I will definitely be experimenting with this at some point in the future. I think I'm often guilty of using too much AA in low res sprites which can have a muddying effect, but I do question why a single pixel of anti-aliasing used to define a shape or curve that would otherwise be impossible at a given resolution can't be considered part of a cluster? A sub-pixel cluster if you like.

Looking through the examples posted in this thread, in particular Cure's, I have to say I prefer the befores to the afters in most cases when viewed at native resolution.  It really seems to me that eliminating the use of single pixels is most beneficial when viewing a piece at a non-native resolution. Given that screen resolutions are becoming higher and higher I can see the value in that, particularly for game art that might be ported to multiple devices with different screen resolutions, but for pieces that serve no functional purpose I question the necessity, and even the validity, of avoiding single pixels.

[rikfuzz]

Seems to cause more problems than it solves (especially round here where the pixel art is already very high calibre).  It looks particularly arbitrary that diagonals have to be double-thick, where other angles don't.  I guess I 'see' lines as a continuous cluster, even though they're broken by diagonals when at 45º or used in curves. 

Still, definitely useful to see where it does improve things, I'm definitely bad for over AAing, and lots of speckles always harms pixel art.

[Pix3M]

... but I do question why a single pixel of anti-aliasing used to define a shape or curve that would otherwise be impossible at a given resolution can't be considered part of a cluster?

I just tell people that we pay more attention to the pixels when they are bigger, and less attention to how everything meshes together into a whole.

[Faceless]

... but I do question why a single pixel of anti-aliasing used to define a shape or curve that would otherwise be impossible at a given resolution can't be considered part of a cluster?

I just tell people that we pay more attention to the pixels when they are bigger, and less attention to how everything meshes together into a whole.

True enough, but I see heavy use of 2x2 pixel squares in Helm's star girl wip, and that is effectively a bigger pixel, but that apparently counts as a healthy cluster by his definition. I'd contend that two diagonal pixels with anti-aliasing to fill the 2x2 square would in most cases look more natural at any resolution than a solid colour 2x2 square. It's particular noticeable in the rib area. Also, I'd say in his effort to avoid using single pixel diagonals he's created more noticeable banding on her waist area, above her hips. Particularly on the left side of the image. It still looks great as a whole, but I'm not seeing how it is improved by a refusal to use single pixels or single pixel diagonal lines. Honestly some of the finished areas kind of look like the second last step in a progress gif before the anti-aliasing is applied.

Another thing I want to bring up is the indents in the door in Cure's mockup. I'd argue that their entire structure (2x2 square hugged by a 3x1 reverse L) is perceived by the eye as one 3x3 square cluster with the shade of the door in two corners acting as single pixel anti-aliasing to make it appear more unified at native res. As soon as you zoom in those areas kind of fall apart. I absolutely think that those indents would look much better when upscaled if they used an extra shade to AA the square corners.

I guess what I'm getting at is I don't think a cluster should always be defined as a group of pixels of the same shade, but rather as a group of pixels that are read as one clump.

---

Edit:
Here's an example of an edit I made (top) to a sprite Slym made that I think illustrates my point about single pixels appearing to be part of a cluster of surrounding pixels that still hold up once upscaled:



Somewhat ironically the edits I made are mostly in line with the principles outlined in this thread before I had even encountered the concept so I seem to have been naturally incorporating it into my work for some time without concious knowledge of what clusters were in pixel terms. The only real change I think I would make after reading this thread is to make the two pixels standing vertically directly above his eye the same shade rather than two.

[Ai]

I guess what I'm getting at is I don't think a cluster should always be defined as a group of pixels of the same shade, but rather as a group of pixels that are read as one clump.

That has pretty much been my definition for a long time: a cluster is a stack of clusters 1-or-more high (usually with brightness increasing as you go up the stack). Inkscape's Trace Bitmap function also implements an algorithm based on this idea and creates highly coherent results.

IME this makes thinking about volume/silhouettes pretty easy. I've never figured out a way to explain it properly that I was happy with, though.

Edit:
Here's an example of an edit I made (top) to a sprite Slym made that I think illustrates my point about single pixels appearing to be part of a cluster of surrounding pixels that still hold up once upscaled:

Yeah, going by the hair, we're on the same page here.
I'd say your edit needs more 'shape flavour' contrast, though (everything seems pretty rounded, whereas there was some round/sharp variation in the original)

[Ryumaru]

Here is my humble contribution to the thread, I hope you don't mind a cross post with my project.



The large rock structure should follow all the rules, the smaller one uses single pixels where I deemed necessary, or just plain wanted them there, and the broken architectural structure is traditional pixel art methods. As a bonus, the main character sprite in the traditional garb should only have single pixels in 45 degree lines. The raven cloaked figures use single pixels relatively sparingly for details or to indicate texture.

To me, the middle ground seems to work the best. If it is convenient, arrange your clusters to have no single pixels, it allows you to have that puzzle fitting power (TM) in some areas, while you still have the freedom to indicate smaller forms or rounded edges with single pixel AA. I like the challenge it brings of thinking of clusters in a different light, and forces me to make choices that I otherwise would have just washed an area in singular AA. Where I don't like it is when I have a clear idea of what I want a form to do, or how I want to make a small area of pixels sing in a certain manner, and the rule is prohibitive. I still am very pro dithering and pro AA, but this has definitely changed how I work with flatter and more planar areas.

[Azuyre]

Decided to try this out by pixeling a plant monster I posted in the sketch thread.


I don't think I have any stray pixels but I did have to use a few 45° connections to get the outline.
Spoiler: Original Sketch

[Basketcase]

I'm loving the results I'm seeing in this thread.

I haven't pixeled much in a while, but this prompted me to try something:

[Helm]

24 megabyte warni-... I guess it's too late now.

I wanted to animate a few bits, but that's as much as I can do for it now. I learned a lot about how photoshop does animation stuff, for this. I'm honestly pretty surprised all those hue shifts fit in a 256 color palette (even with dithering on).

[Mr. Fahrenheit]

Really cool helm!

I can't help but feel that on some of the smaller detailed things using some aa and single pixel would've increased the readability  . The hands mainly.

Awesome picture though and I always love the subtle animations you use on some of your pictures. Speaking of the animation, did you stick to the no single pixel rule as you animated as well or did you erase the extra pixel if it intersected with something?

[Helm]

Nope, there's too much parallax to check every frame. This is 900 individual frames for a 30 second loop. Thanks for the kind words, btw

[PixelPiledriver]

Thats really cool Helm. 
The stars jump every 4th/5th frame to reach their destination.
It might help to do the math for a smoother movement.

[Helm]

m-m-- math?!

[Joe]

Fantastic thread. I've been looking at it for a long time, today I finally set out to try it.  I feel the most appropriate name for this process is defragmentation.



To the best of my ability, there is no banding or orphan pixels. If there are, I was sloppy and missed it. I used 45 relationships sparingly and usually as a "cheat" to adhere to the restriction but still single pixel AA.

I found moreso than ever, I was catering to the pixels themselves rather than the drawing. That is, it's not as accurate as a painting would be, because (my own personal inaccuracy aside) instead of representing the source faithfully I had to bend to the will of the clusters.

Particularly difficult was avoiding banding. In a few cases (example: right eyelid) it was pretty much impossible to avoid; I had pixelled myself into a corner and the only choice was which two tones to band. The combination of no banding, no dithering, and no single pixels was very challenging. I will continue to experiment with this idea since one piece is quite insufficient, but nonetheless it gave me many questions.

1. Banding is probably the most incompatible with defragmentation. In many cases I had to contrive the subject, and especially the details, to avoid it and adhere to the restriction. My question: which, in your mind, takes precedence between the three? Banding, Defragmentation, or Fidelity? I have a feeling you'd say it's on a case by case basis. My personal tendency is to prioritize banding-elimination.

2. Why does this work? What is it about the absence of orphan clusters that makes good pixel art?

3. AA was discussed earlier in the thread, as something that is less necessary. Wouldn't sharper screens make AA more necessary than before? Or did I misunderstand that point

I'd have to practice a lot to see what that means intuitively for my own art

4. After a year of practice, where do the single and double pixel stand in your mind? As I understand it this exercise is meant to strengthen the integrity of your clusters, so that the single pixel becomes a tool, not a crutch. But there were many places where I knew single pixel AA would be the best solution, I merely refrained. I'm sure you don't think single pixels should be abolished; where do you feel they're appropriate?

E: Found more single pixels.

[Ai]

Fantastic thread. I've been looking at it for a long time, today I finally set out to try it.  I feel the most appropriate name for this process is defragmentation.

That's an excellent name for the pixel art version of this.
For more CG-ish stuff, there are filters like GMIC's 'anisotropic smoothing', whose results, IMO, express the artistic ideal behind defragmentation -- everything being 'well fitted' to everything around it, and also show how not to apply it / when to break the rules (fine details becoming too indistinct). I mention that because they use a term in that that is also very fitting for the process we're discussing here, 'regularization'. It is more general than this specific process but captures the reasons for applying it.

I found moreso than ever, I was catering to the pixels themselves rather than the drawing. That is, it's not as accurate as a painting would be, because (my own personal inaccuracy aside) instead of representing the source faithfully I had to bend to the will of the clusters.

Of course, you could still bend more. Arguably the logical extension of these principles is to flatten planes in order to minimize cluster contention, ending up with something that is quite vector-y, which perhaps you could have used on the lips; I perceive them as more detailed than everything else.

Some of the stuff in ptoing's PixelJoint gallery demonstrates other ways of taking it further -- like exaggerating planes and faceting so that they are easier to fit together.

1. Banding is probably the most incompatible with defragmentation. In many cases I had to contrive the subject, and especially the details, to avoid it and adhere to the restriction. My question: which, in your mind, takes precedence between the three? Banding, Defragmentation, or Fidelity? I have a feeling you'd say it's on a case by case basis. My personal tendency is to prioritize banding-elimination.

I have the same tendency, but I am willing to confidently state that Defragmentation > Banding, due to mentally ranking them by how involved they are in making the overall work hang together. Fidelity is really a different type of thing IMO, fidelity is a rather selective thing in good art IMO -- you decide what aspects are important to capture and then do so. Fidelity to aspects not on that list is mostly a distraction until you reach the stages of seriously polishing the work.

2. Why does this work? What is it about the absence of orphan clusters that makes good pixel art?

Personally I believe 'regularization' addresses this question.

3. AA was discussed earlier in the thread, as something that is less necessary. Wouldn't sharper screens make AA more necessary than before? Or did I misunderstand that point

As I understand it this is mainly a question of PPI. The more pixels-per-inch our screens have, the less we need to use kludges like AA (sacrificing colors for the impression of greater resolution). With sharper screens, the pixels in AA are more apparent, but with higher-pixel-density screen, each individual pixel is less apparent (so, AA pixels are both less obvious and less effective). When pixel art was really big, we had PPI as low at 30-50. Now, 72 PPI is about the minimum you will see, 90-96 is commonest, and we are looking at even higher densities.

Retina displays probably need separate consideration, but what I say holds for conventional display devices (phones, tablets, laptops, desktop computers)

E: Found more single pixels.

If you use GrafX2, I made a script that detects them earlier in the thread.

[Joe]

Thanks for the well-thought reply Ai.

I'm sure there are better names to describe it, I think you got really close with 'noise reduction,' cause that's what it is essentially. Moreso probably than defragmentation—I was thinking of how all the stray pixels were being absorbed into their parent clusters, kind of like how data gets reorganized, but that isn't quite what's going on. I think noise reduction is a more accurate description and yes, this is definitely in the same vein as anisotropic smoothing. Culling and truncating also came to mind. I am aware of the caution toward inventing terms here, just trying to summarize this idea in my own mind, as well as distill it into a word.

It is more general than this specific process but captures the reasons for applying it.

I looked up regularization and found

In simple terms, regularization is tuning or selecting the preferred level of model complexity so your models are better at predicting (generalizing). If you don't do this your models may be too complex and overfit or too simple and underfit, either way giving poor predictions.

So if I'm getting this right, this is a less complex model, because it doesn't account for information that single clusters would. Or is it more complex because it has more rules?

Personally I believe 'regularization' addresses this question.

I tried to make this connection, but I don't understand it well enough yet. How does regularization explain why orphan elimination makes clusters stronger?

Yes you're right, fidelity is selective. I guess coming from a traditional background I'm so used to it being the point, but I'm starting to see how it's not the point with pixels. What you're saying is defragmenting is even more the point than banding elimination, that banding is almost accessory. I think I can see that.

Arguably the logical extension of these principles is to flatten planes in order to minimize cluster contention

Yes, if you were to take it a step further it would then be a matter of reducing the clusters themselves (like with the lips). Interesting... this makes a sort of gradient, then, with perhaps dithering on one extreme and large, flat planes on the other. Similar to how banding has its own spectrum of very obvious to less obvious. So then, how you find balance between the two spectrums, I think that would be a useful discussion.



Yes that's a very clever script, thanks for making it. I will check out Grafx2, especially since it has a linux build. Oh, also

With sharper screens, the pixels in AA are more apparent, but with higher-pixel-density screen, each individual pixel is less apparent.

This I follow, but I guess I had the underlying assumption that people aren't viewing pixel art at 1x on our modern screens. I always thought that pixel art was designed for smaller resolutions, even today where the games that use it either scale it up or have a low native resolution, which would make AA more important, since screens are indeed sharper.

[Ai]

In simple terms, regularization is tuning or selecting the preferred level of model complexity so your models are better at predicting (generalizing). If you don't do this your models may be too complex and overfit or too simple and underfit, either way giving poor predictions.

So if I'm getting this right, this is a less complex model, because it doesn't account for information that single clusters would. Or is it more complex because it has more rules?

Hmm. It helps to understand spline math, or gaussian blurring, I guess. There is the concept of quality of fit, but more important in this case, there is the idea of how well fitted things are to the things surrounding them.
For example, a conic spline is fitted to 3 variables, a cubic spline is fitted to 4.. In practical terms, fitting features to other features tends to mean that the features 'interlock' tightly -- many points on each feature align (along a line or more commonly a curve) with many points on other features, producing an image that feels very integrated and solid. The anisotropic smoothing filter I brought up tends to move an image towards this ideal (with the obvious downside that you may not want -all- features equally regularized.)

So I would say that this is a model that fits more variables, but also intentionally discards many variables (eg. those that you might find in trying for precise fidelity to reference), to create a whole that is much more obviously unified and beautiful (in a literal sense, but also perhaps in a mathematical sense)

Yes you're right, fidelity is selective. I guess coming from a traditional background I'm so used to it being the point, but I'm starting to see how it's not the point with pixels.

What you're saying is defragmenting is even more the point than banding elimination, that banding is almost accessory. I think I can see that.

Yeah, I like to kill banding, but realistically, it adds maybe 2-5% rendering quality, and the conclusion I think we've more or less agreed on in this thread is that solid clusters constitute 70-80% rendering quality -- which is enough to make a picture hang together, and coming back and doing debanding and AA is something that can go into a separate pass without noticable influence on how the assets work.

Yes, if you were to take it a step further it would then be a matter of reducing the clusters themselves (like with the lips). Interesting... this makes a sort of gradient, then, with perhaps dithering on one extreme and large, flat planes on the other. Similar to how banding has its own spectrum of very obvious to less obvious. So then, how you find balance between the two spectrums, I think that would be a useful discussion.

I'm not sure what to contribute to that discussion, but I agree.

With sharper screens, the pixels in AA are more apparent, but with higher-pixel-density screen, each individual pixel is less apparent.

This I follow, but I guess I had the underlying assumption that people aren't viewing pixel art at 1x on our modern screens. I always thought that pixel art was designed for smaller resolutions, even today where the games that use it either scale it up or have a low native resolution, which would make AA more important, since screens are indeed sharper.

Hmm. I may be a bit out of touch here. I am aware that some games scale up their art, but I had no idea how prevalent this was.
I would suggest that the effect on scaled up pixel art is slightly different, more like a simulation of gaussian blur, from my own subjective experience; it imparts a much more ambiguous type of smoothness the greater the effective pixel size. I find it adds something, but a different something than AA on lower-res / blurrier displays.

It's good to revive this discussion, hopefully others will get involved again soon.

[Helm]

First of all, lovely art you've got there, Joe. I can look at it and immediately see how the process has helped you.

As to answering your questions,

Wwhich, in your mind, takes precedence between the three? Banding, Defragmentation, or Fidelity? I have a feeling you'd say it's on a case by case basis. My personal tendency is to prioritize banding-elimination.

Removal of banding gets a bit metaphysical at some point, where just by the process of stripping it out I often feel I am arriving at something more robust, but robust in what sense I can't really tell you, it's not making the subject clearer only, or adding detail (often it is removing detail). The way I would liken it is that zen quality of getting something down in painting with few colors and few brush strokes, it's not even exactly 'control' what I am describing here, more like chiselling something until it arrives at some quintessential level. Of course I can't explain this much better, suffice to say I've been working professionally, using most of my techniques (more on this later) to the point where they're second nature and they never interfere with getting things done to the best of my capacity, they only give an added self-sustaining enjoyment of getting things down to that quintessential level.

Defragmentation as you describe it I sometimes fight against consciously, adding more shades and blurs and dithers when the subject matter seems to desire it although that creates more softness and complication where a clear cut cluster could best suit. It isn't very often that this happens, though.

Fidelity, honestly for small-ish arcade game sprites, single pixels will happen, but aside from that I've never felt that I am sacrificing fidelity by banding removal and cluster clear up. The more you practice, the more it'll become second nature, there are wonderful solutions for most problems, and then there's that 'quintessence kick' when you find them.

2. Why does this work? What is it about the absence of orphan clusters that makes good pixel art?

I think it's flow, two (or more) married pixels always have a direction, as a brush stroke. An orphan pixel is a directionless dot, it has its uses but it doesn't suggest motion or connection between parts. When the contrast of the piece is such that you can both see the connections of clusters but also see them as related by lightness and shade, then you have a very harmonious, moving, sculpted entity.

3. AA was discussed earlier in the thread, as something that is less necessary. Wouldn't sharper screens make AA more necessary than before? Or did I misunderstand that point

No because AA relies on blurry crts to buffer a color in a color. The end result is that the blurry CRT art looks like painted art, convincingly, not that it looks like good, crisp pixel art. When you look at pixel art on a sharp screen, you're never going to hide the fact that the art is made of pixels as effectively as on a bleeding CRT monitor, so why try? Show pixel lines (esp, perfect lines), show connections, keep it crisp.  Would a game of tetris look more beautiful if you ran it through a gaussian blur filter? Digital art gets to be super crisp, this is an asset, why treat it as a deficit?

4. After a year of practice, where do the single and double pixel stand in your mind? As I understand it this exercise is meant to strengthen the integrity of your clusters, so that the single pixel becomes a tool, not a crutch. But there were many places where I knew single pixel AA would be the best solution, I merely refrained. I'm sure you don't think single pixels should be abolished; where do you feel they're appropriate?

I don't adhere to a no-single pixel rule in my daily professional artwork. It's just not practical for very 1. low-res artwork that is also 2. very cartoony stuff which is what I'm employed to do at the moment. I do try to minimize single pixel usage unless it's highlights, dot eyes, that sort of thing. Aside from that, my point of view is largely unchanged from when I last wrote on this thread: banding removal/defragmentation of shapes is the bedrock of my  technique and I think it's the only honest piece of craftsmanship that should be communicated and practiced for the learning pixel artist. The rest is just fluff that we created (for good or worse) in a community bent on classification and enumeration of 'techniques'. This is very common, most creative online communities create jargon. I do not return to 'selout', for example, as an effective and intersectional technique but I do think about banding and pixel shapes every day, in my work.

Tangentially: my recent interest is in bit restrictions for palettes, esp. 3 bits per channel, like the Atari ST. I use this restriction in my daily work because it explains the color choices of past videogame artwork very effectively, a lot of linear from-hue-to-different hue ramps, full saturation, less use of gray, less use of interchangeable colors, a stronger foundation in pure black for segmentation of forms and so on. This is a blind spot in how we look at art in Pixelation, because we jumped into learning pixels with a 24bit palette in our hands and no matter how restricted and small the palette we used was, it was still picked from a color space of millions of colors. So many close shades, so many earthly greys, so on. It is a joy for me to un-learn, for a time at least, my dependence on shifting colors through grey buffers to get things done, and it's really no way to practice this with the c64 palette or the EGA palette, honestly. I encourage anyone with the means to restrict their bits-per-channel to give this a good go, especially with 3,3,3 and 4,4,4.

[ptoing]

The 333 and 444 thing is very true, fully agree there. Moreso 333. 444 is a good step closer to super smooth, esp if you have as many colours in your picture as you want. 555 is where you have more colours than you could ever want for pixelart in most cases.

I personally also have doen a lot of RGB222 (full 64 colour EGA palette, Sega Master System for example) stuff which is good fun.
It forces you to make bold choices with your colour usage.

About AA making less of a point on modern displays I only agree partly. It depends at what magnification you are looking at the work.
If you are making an application with pixeled icons that are looked at at x1 then yes, AA helps for sure. And depending on how high the PPI on your display is even at x2 and x3 the aa will hold up and soften things (esp curves), but I agree that it is pointless to try and make everything super smooth, or what I would say turns into overAAing things, which is something you can for sure get away more on a blurry CRT. But at the same time on a CRT you DO need less AA to get to the same level of smoothness than on a LCD.

[Joe]

@Ai: That makes more sense to me, thanks.

@Helm: Thanks! And yes, it really clicked when I tried it.

Stroke direction, that's probably it. I also see that single pixels are noise in most places. It definitely feels brushy when the clusters are aligned.

Obviously this isn't the only way to pixel, much like paint has many ways of being applied. I see some artists like tomic pull off dither very convincingly even on my crisp screen, and I would like to add that to my toolkit. But I am convinced this is the best way to pixel at present, barring any stylistic intentions.

The use of AA—like dither—is subjective, I don't see any conclusive evidence that it has less of a place. I say this with the assumption that over-AA was never appropriate. Here is an arbitrary curve:



Clearly, the buffer color is there; it's not fooling anyone. But it's a trade-off, and I think the benefit outweighs the cost, that the curve is effectively smoothed. Yes you could convert the curve to perfect lines, but that's not always an appropriate solution. That said, thanks for clearing up why LCDs reduce the effectiveness of AA.

On bit restrictions: That's been mentioned here before, but I haven't seen it discussed or described in depth. If you feel strongly about its usefulness for learning, why not start a ramblethread in that vein? I want to try but don't know where to start. Also why would bit restriction exclude grays? Or is that just personal choice

@ptoing: Are there any introductory articles you'd recommend on the subject? Don't know how you would calculate whether a color was in that space, how you would set up the palette, etc. Feel like something that fundamental should be in the linkage thread.

I revisited the Altered Beast sprite with interesting results. Less cautious and more daring is what comes to mind of how I felt while working. There were many solutions I never would've considered before. Original -> Old -> New:

[ptoing]

Joe: There are 2.5 approaches as far as I know.

Unstretched:
256 / (number of steps per channel). So for example for 444 you would have 16 steps.
Then you always add that number to get your 16 values. Which means you do not get to the full range of 255 but in this case you stop at 240 or F0. And your legal values for 444 unstretched would be 00, 10, 20, 30 and so on.

Stretched:

This means you expand to the full range of RGB888:

To do this you go: 255/steps-1
This is your base step, and then you multiply that by whatever step you want to get at. So for example if you would take 5 bits, being 32 steps you have 255/31 = 8.225806..... and you wanna know the 8th step you get 65.8064.....

Now this is why I said 2.5 ways. Because when stretching you can either floor (aka always round down) which would give you 65 in this case, or you could round to nearest so that would make it 66.

In Promotion you can set your bits per channel, and in Grafx2 you can set steps. Both have their pros and cons.
With bpc you can do stuff like 232 or 565, with steps you can do 3 steps for which there is no clean bit value. But the Amstrad CPC for example had 3 steps per channel resulting in 27 possible colours.

But if you know how to calculate it it is fairly easy to make small palettes like the CPC one by hand and save them.

I hope this was clear enough, if not let me know.

[Ai]

On bit restrictions: Also why would bit restriction exclude grays?

Well, look at an example, Amstrad CPC:

You have three possible levels of R,G,B : 0, 127, 255. That gets you 3 pure gray levels : black, midgray, white.
There are a few mildly desaturated colors, like 255 255 127, but these certainly are nowhere near gray. For mixing colors in the way helm is talking about, you need relatively desaturated colors. CPC art mostly avoids creative color mixing for this reason.

As you increase the number of RGB levels, more desaturated colors become available: For example, in EGA64, you have 4 rgb levels: 0, 85, 170, 255.
This gives 4 pure greys plus a few 'sort-of-vaguely-grayish-if-you-squint' colors like 170,170,255.

.. by the time you get up to 16 levels / 4 bpc (Amiga), you have a large choice of hues, 16 pure greys and 30 or so off-grays and can definitely do plenty of blending. The majority of colors available are still relatively saturated, though -- I'd say for example that there is not a 'non-oily' caucasian skin tone in a 444 palette -- you have a choice between rather olive and a bit red in the face.


In addition to what Ptoing explained, it's also rather doable to emulate color restriction, if you have to, via hex #RRGGBB codes.
For example, allowing only one of 00 40 c0 f0 for each channel is equivalent to unstretched ega64 -- 2 bpc. amiga / 4 bpc is also easy to remember, it's just 00 11 22 33 etc up to ff (or #00 10 20...f0 if you're going for unstretched).

[ptoing]

Well, not 16 greys, always steps minus 2 since black and white are not really greys and can not be used to buffer anything either.

[Helm]

re: restricted bit palettes for learning: not really, honestly. I would recommend people work with 8bit system specs, of course, but it's not like you can't learn with the c64 palette or the EGA one.  It's more useful for me because I went through a decade of amiga-esque blending and I'm trying to unlearn certain habits and also love game art for being game-arty.

On a more general note, I don't want to add more to my thoughts on cluster theory even if I had something significant to say because even with the last major concept I introduced (the examination of single pixels) there was significant pushback. I feel that there's no real need for a single person to push this envelope when a community says 'uh... nope'. Different people, in the future, fate willing, will contribute more palatable ideas that build on this foundation. The way the cluster stuff has been recombined and streamlined by Cure for example pleases me greatly - as divorced as it can be from any single person's ego, the better.

[ptoing]

Yes, I think everyone should try and take away their own lesson from what cluster theory can do for them. I find myself being a lot more conscious of cluster formations and things like single pixels, but I still use them where I feel they are warranted.

I think it shows quite nicely in my latest bigger piece, which coincidently also is RGB333, since it is Atari ST specs.

[yrizoud]

MCGA screen mode (often called "VGA" 320x200x256) seems to have a precision of only R6G6B6. (64 levels of R G B)
Is there somebody who worked with this mode, who can say if this felt like a limitation in color choice ?

[Ai]

MCGA screen mode (often called "VGA" 320x200x256) seems to have a precision of only R6G6B6. (64 levels of R G B)
Is there somebody who worked with this mode, who can say if this felt like a limitation in color choice ?

Not for color choice generally, but desaturated colors are still a bit lacking -- it was an issue that came up when I was designing the new default palette for OHRRPGCE. I'd go up to about 192 levels of R G B before I'm -really- satisfied with choice of desaturated colors.
It's a pretty mild limit though -- amiga 444 encourages you to be obviously oilpainty in color usage, VGA is just slightly oilpainty, and won't really be a relevant concern for < 10color ramps.

[Helm]

Worth keeping in mind that stellar VGA work did not occur in a 666 bit vacuum suddenly in 1990 or something. The artists that did the best VGA work were trained on previous 8bit machines and more importantly on amiga, on deluxe paint and brilliance and so on, therefore their color choice and technique does not rely on infini-shades and 256 full palette pictures very often. Look at Westwood artwork in particular, most of that stuff happened on amiga first and then got dos ports.

[ptoing]

MCGA screen mode (often called "VGA" 320x200x256) seems to have a precision of only R6G6B6. (64 levels of R G B)
Is there somebody who worked with this mode, who can say if this felt like a limitation in color choice ?

I worked a lot in 555 for GBA and DS stuff, and I never felt like that was a limitation at all.

[Helm]

I agree. 555 might as well be 24bit, as far as pixel artistry goes.

[Arne]

Even The Amiga's 444 isn't much of a hinderance unless you're doing copper stuff or long 32 color mode ramps. I guess, if the gradient is so smooth you can't see the transitions, it stops being pixel art anyways because there is no pixel definition.  I do recall some instances where a hue shift "one step up" in a ramp has been unwanted though. Some games would offset the copper half a bar every other frame to smoothen out gradients.

As for going lower, it's interesting how such stepping produces colors almost the opposite of what we see on photos where the most common colors stretches from the black corner of the color cube to the white, diagonally through the grays bulging like a grain of rice.

On topic, I've found clustering/orphan elimation a useful thing to fall back on when something feels wrong but I'm somehow blind to it. It's good to have there at the back of the head when the old orges rise to do uneccessary fancy pixel footwork with excessive AA, lining and whatnot.

[Ai]

Even The Amiga's 444 isn't much of a hinderance unless you're doing copper stuff or long 32 color mode ramps. I guess, if the gradient is so smooth you can't see the transitions, it stops being pixel art anyways because there is no pixel definition.  I do recall some instances where a hue shift "one step up" in a ramp has been unwanted though.

Yeah, for individual sprites it's no problem and even helpful, for long gradients like in a 256c master palette, it messes around with saturation and hue when you go to fill in intermediate colors. (basically why I designed the new OHRRPGCE palette with amiga 444, then went to VGA 666 and eventually full 888 when it became available, to get easier-to-work with midcolors. The result had to be usable and understandable by people without much understanding of pixel art or color theory, and it had to have long (15/16c) ramps.)

As for going lower, it's interesting how such stepping produces colors almost the opposite of what we see on photos where the most common colors stretches from the black corner of the color cube to the white, diagonally through the grays bulging like a grain of rice.

I'm not sure what the opposite of 'the most common colors stretches from the black corner of the color cube to the white, diagonally through the grays bulging like a grain of rice.' would be. Are you proposing a concave rather than convex shape occurs, wide at black and white and narrow in the middle, like an hourglass with the outside corners bevelled flat?

[Arne]

Well, some years ago I gathered a bazillion photos of a variety of things (including naked people ) and plotted a 3D RGB histogram of color count. Most hits were aligned from the black diagonal to the white, bulging at the center (midpoint) like a grain of rice... perhaps one with fuzzy irregular mold. Certain saturated colors were of course less common as they are less likely to occur in a photo. A palette generated by nested RGB for-loops using big steps produces few hits on the common color regions in my aforementioned histogram. Of course, graphical neon colors are still quite usable, perhaps more effective in many cases. Would be interesting to try other stepping methods though, for say a 64 color pal, and find one that's more to my liking whilst being really simple.

[Ai]

Ah. so what you're saying is that retro colorcube is roughly the boolean subtraction of the photo profile from the total RGB colorcube. Something very roughly like this:


Where white is occupied locations.

I've tried different stepping methods myself, and different colorspaces. YCbCr is a moderate improvement on RGB. Colorspaces that cannot be expressed with a plain 3x3 transform matrix, but nonetheless are carefully fitted into sRGB, may be an improvement on that -- eg the variable-precision/dependent H/S/L encoding used in png16.

[32]

I realised after reading this that I'd kind of been doing the channel restriction thing already just cause it was satisfying to me, the colours felt more pure or something. It hadn't occurred to me that it had any basis in history. Having been told that I ran with it on my newest piece and I think the colours turned out really nice. Of course I went with 5 values per channel which makes no sense in terms of bits but it came up with a really interesting palette. Does anyone smarter than me know of a way to generate the whole palette with those restrictions?

Between this, banding and clusters I think every time I read a new theory from you Helm I get better Just wanted to say thanks.

[Ai]

I realised after reading this that I'd kind of been doing the channel restriction thing already just cause it was satisfying to me, the colours felt more pure or something. It hadn't occurred to me that it had any basis in history. Having been told that I ran with it on my newest piece and I think the colours turned out really nice. Of course I went with 5 values per channel which makes no sense in terms of bits but it came up with a really interesting palette. Does anyone smarter than me know of a way to generate the whole palette with those restrictions?

In Gimp, you can generate RGB colorcubes using 3 layers, a bit of pencil tool work, a bit of scaling, the filters->Map->Tile filter (or copy/paste), and Colors->Components->Compose filter. Just keep in mind that the image must be Grayscale, not RGB, for the Compose filter to be available.

These are the three grayscale layers I made, they each represent a channel -- order doesn't matter but I like having green going across, red going down and blue as z (dunno what else to call it -- the slowest-changing dimension), so that is how I've labelled them here:

R:
G:
B:

(given the above layers, all you'd need to do was chuck them in a Greyscale image and feed them to Colors->Components->Compose. The other comments described how I created the above layers. Hopefully the structure of the layers is obvious.)

These images are 5x25, or more generally Ax(BxC). Which dimension is which doesn't matter, 25x5 would be equally doable.


If you want values you can use in code, it would be pretty easy in python, just use

Code: [Select]

import itertools
alltriples = itertools.product(rintensities, gintensities, bintensities).
for example for your 5x5x5 cube you can use the values you mentioned:

Code: [Select]

import itertools
intensities = [0, 0x3f, 0x7f, 0xbf, 0xff]
alltriples = itertools.product(intensities, intensities, intensities)
Both methods are easily adapted to arbitrary intensity levels or different number of intensities for different channels.

Though of course as Ptoing has mentioned, GrafX2 or ProMotion are probably more convenient than generating a palette, since they explicitly force all colors to conform to the selected colorcube at all times.

The 5x5x5 palette doesn't seem to be used historically. OTOH, the Amstrad CPC palette was 3x3x3, and 5x5x5 is a direct expansion of it (keeping three levels exactly as is, adding two exactly intermediate ones). So I mentally labelled this colorcube 'CPC5'.

[32]

Of course! Now that you've told me it's so obvious . Very cool tool in gimp also, tried doing it in photoshop but I guess all channels are not created equal or something according to adobe. Definitely want to get my hands on a copy of ProMotion but the Australian Dollar is so weak right now I can't quiiiiite justify the expense. I remember seeing it used years ago in the recorded pixel art thread though and the colour handling looks far superior to GraphicsGale. Between that and the tiling tools I might cave on it soon.

Interesting about the CPC. I obviously don't have all the relevant knowledge but I'd assumed odd numbered palettes wouldn't exist.

[Ai]

Of course! Now that you've told me it's so obvious . Very cool tool in gimp also, tried doing it in photoshop but I guess all channels are not created equal or something according to adobe.

Yeah, pretty sure you can do it in photoshop, but no idea how.

Definitely want to get my hands on a copy of ProMotion but the Australian Dollar is so weak right now I can't quiiiiite justify the expense. I remember seeing it used years ago in the recorded pixel art thread though and the colour handling looks far superior to GraphicsGale. Between that and the tiling tools I might cave on it soon.

Interesting about the CPC. I obviously don't have all the relevant knowledge but I'd assumed odd numbered palettes wouldn't exist.

Well, that is with a <= 16-color display (ie. you can pick 16 of the 27 for your usable palette), so the actual number of bits per pixel is still whole.

FWIW I messed around with RGB curves, producing this modified 5x5x5 "cube" that IMO has better differentiation in red and blue:


(blue ramp is 00 5d a7 c8 ff, red ramp is 00 56 87 ae ff. Pure greys are sacrificed due to this, you get tinted greys instead.)

Unadjusted version:



Comparison unadjusted | adjusted:


IMO it's noticably less 'overexposed' looking.

[32]

I'd assumed I could just drop the images into the channel layers photoshop has. At first impression I'd thought it worked but for some reason the channels weren't mixed properly .

Probably could use some messing around, I think if anything I'd drop the first value DOWN though cause the palette lacks much in the darker end. I tried applying your palette 1 for 1 to my image but I didn't really gain any insight, it looks more over exposed but it's impossible to say without manually re-picking the colours. I kinda like the tinted greys though and eyeballing it, it does look a bit nicer.

Do the values of the steps on the old machine palettes have any basis in the hardware? Is it kind of interpreted as uniform steps of brightness for the subpixel (I don't know much about bit depth in terms of old monitor tech haha)? Obviously there's no reason not to play with them now but I'm curious.

EDIT: Thinking about it I guess the values of the steps would just be determined directly by the limitations of the monitor technology? So it is kinda of arbitrary in a way?

[Ai]

Yeah, the adjustment I made has a side effect of reducing dark colors, which is an intuitive judgement based on experience drawing with EGA64, CPC, and unaltered RGB555 palettes. The primary effect I was aiming to achieve was turning all the similar oversaturated colors into  colors with more varied saturation.. so more pastel / watercolor aesthetic, I guess. Done using a very mild, oblique S curve (exact shape differing for each channel).

The values were often determined directly by monitor technology, yes. Though monitor technology quickly began moving in the general direction of what we use today, sRGB, so most of them can be fitted pretty accurately / simply into sRGB.
Ptoing's limitations thread goes into it more, but in general there were two types of possible ramp, clipped and unclipped.
Unclipped, the step value is (255/(N-1)). Clipped, the step value is (256 / N) (meaning you don't get "full" intensity -- eg. for N=8, which Genesis used, the maximum intensity is 224 aka (256 - (256/N)) ).

It's not uniform steps in brightness (for RGB based systems), it's uniform steps in something approximating sRGB space (which includes gamma companding of approximately 2.2, so the response to intensity values is exponential  rather than linear.)
I wrote a bit more about that here, if you are interested, but the short story is it tends to work out pretty well for small colorcubes -- significantly more pop/expressiveness -- to just pretend that sRGB is linear even though it's not (so, eg. 00 ba ff does not look as good for CPC 3x3x3 cube as the 'naive' 00 7f ff, even though 186 is perceptually more 'correct' as a midpoint).

FWIW if you want to see what an semi linear colorcube looks like you can apply gamma 0.45 (1/2.2) using Levels tool. It makes all the classic failings of small colorcubes significantly worse

[32]

Very interesting stuff, there's clearly a lot I don't know but have observed which will be nice to back up with some solid information . I see what you mean, almost all of the colours I was using were completely saturated, which I liked but the lack of variation could be a hindrance, especially if I was using a larger selection. Ptoing's thread went over my head at the time but I'm definitely gonna go back and have a look at all this stuff, see if I can wrap my mind around it. Far more valuable knowledge than I'd initially given it credit for I think. Thanks Ai 

[Helm]

32, I appreciate your kind words. Your pixels are beautiful, by the way 

[ptoing]

The colourcube thing that Ai described you could also do pretty easy in Photoshop by directly drawing into the RGB channels.

[32]

Thanks Helm, I'm sure you can see your influence all over it, that rock is just for you

That's what I tried Ptoing but it completely didn't work. The values got all messed up when they mixed for some reason. I tried turning off colour handling and it didn't help so I gave up the investigation since I already knew the tool worked in Gimp. Maybe I'll look into it again today cause it was pretty confusing.

[Helm]

If I ever have to work on a platform game with water in the future I'll try to do that trick with a ripple being triggered just by going near the edge of the water, really cool.

[ptoing]

It definitely works, you must have done something wrong. Basically you have to make sure that when you go into the channels tab to turn off all the channels but the one you are drawing in. What is best to start out is to draw the greyscale values you are using in the normal layer, so they will be in all 3 channels and are easy to pick from.

Here is a PSD I just made.

https://dl.dropboxusercontent.com/u/15588722/post/32colorcube.psd

[32]

I tried it again just now and it worked fine. Actually turns out it's cause I just copy pasted Ai's images into the channels rather than drawing them by hand and photoshop modified the values for whatever reason. Probably should've caught that one 

Edit: Missed your post there Helm. Yes I think it would be particularly fun. One of my favourite touches ever was in The Last of Us where you can push the office chairs to spin in circles. Something satisfying about a game not being as rigid as you thought it would be.

[ptoing]

Yay It seems like a faster way to generate arbitrary palettes than the way Ai described.

[Ai]

^^ Yeah, I just tested the directly-drawing-channels thing and it works in GIMP, but pasting stuff in affects all channels at once. Sounds like PS and GIMP are doing things very similarly in this case.

^ Eh, if you think my method is substantially slower than yours, you probably haven't realized that it's basically the same method, just without realtime preview, and with the ability to work with more colorspaces. eg. YCbCr is one that's also relevant to retro colorspaces (NES, C64.. Atari IIRC).
Example YCbCr 3*** colorcube:

[surt]

This motivated me to start that colour cube generator I've been meaning to make for ages.
http://img.uninhabitant.com/colourcube.html
Level mode and custom levels aren't implemented yet.
Single file HTML and client-side JS only, so easy to download and run locally.

Any suggestions?

[Ai]

I was surprised that "Level mode (stretched, clipped)" did not effect the result of Bits per channel -- because basically all clipped colorspaces defined by existing systems are expressible in bits per channel. I realize that you could say 'well, use levels instead for that' (once levels is implemented), but IMO it is less surprising if that particular option applies universally.  Maybe instead of 'level mode' you could call the option 'intensity mapping'.

... Or you could automatically sync the bits / levels fields where possible (eg bits is set to 3 when user inputs levels . That would be simpler and more functional maybe.


(.. and I just now realized that Level mode was supposed to specifically be a part of the Levels section. That's a formatting issue -- My quick-and-dirty html skills would probably have stuck it in a blockquote to clarify its subordinateness)

[ptoing]

Ai: Hm, the 3D cube looks like CPC colours and the small cube does not. Are they supposed to be the same? Is there some weird colou profile active on the small png?

[RAV]

I worked a lot in 555 for GBA and DS stuff, and I never felt like that was a limitation at all.

I agree. 555 might as well be 24bit, as far as pixel artistry goes.

As a funny anecdote:
My old Afterlife demo basically represents all of 555 colour space to explore, that is 32x32x32 rgb-cubes. (the whole scene is 64x64x64, but that is just symmetric repetition.)
I've still kept this as a mode in the new to collect a palette. The space can switch between representing 111 - 555, and it just hasn't seemed necessary to go further, though it could.
This is just an optional helper map-mode for starters though. Once picked, colours can still be fine-tuned in full 256 steps by their standard rgb-bars.

[Ai]

Ai: Hm, the 3D cube looks like CPC colours and the small cube does not. Are they supposed to be the same? Is there some weird colou profile active on the small png?

Hmm? I don't think I have posted a CPC cube in this thread. The only 3x3x3 I posted was the YCbCr cube in my recent post, which AFAIK isn't a master palette used by any real system.
Unless by 'small cube' you are referring to the content of the large 3d cube image I posted in the OT thread -- there are three or four levels of compositing there and I didn't quantize the result back to CPC colordepth.

Looking at the stuff I did when messing around with colorcubes, if you want an adjusted CPC colorcube there is this (less extreme saturation, each color is more distinct). I hadn't linked it here before.

[surt]

Updated: http://img.uninhabitant.com/colourcube.html
Implemented custom levels and intensity mapping (I like that) options.
Auto update not implemented yet.

I was surprised that "Level mode (stretched, clipped)" did not effect the result of Bits per channel

I'm not sure what you mean. They only effect how the number of levels (which can be calculated from a bit depth) are evaluated at 8-bit per channel.

(.. and I just now realized that Level mode was supposed to specifically be a part of the Levels section. That's a formatting issue -- My quick-and-dirty html skills would probably have stuck it in a blockquote to clarify its subordinateness)

Hopefully the current formatting is clearer.

[ptoing]

Levels per channel seems to be quite fucked if you do weird numbers, like 2, 3, 255. The output makes no sense at all.
Also you should be a able to have 256 steps not just 255. 0-255 = 256 steps.

[Ai]

EDIT2: BTW surt, this report is still accurate after your posted fix for ptoing's error below.

I think I found a bug -- entered 3 in the middle field, left others at 2. In the link, you can see that the result is clipped.. and also there don't appear to be 2**3 == 8 distinct Green levels, rather, there are still 4 distinct levels and they are used .. improperly? Well, the result doesn't quite make sense when you look at it, that's for sure.

testing some other values:

2, 3, 3 looks correct
2, 2, 3 does not
3, 2, 2 looks correct
1, 2, 3 does not
3, 2, 1 does not
3, 1, 2 does not
2, 3, 1 does not (wow, very not)
2, 1, 3 does not

I've looked over the code, but nothing jumped out at me as being wrong in a way that would cause this effect.
(also, your code has caused my opinion of JavaScript to rise. Doing that in that little code? It ain't NumPy, but it's pretty good.)

FWIW, the levels readout always appears correct, so those arrays must be populated correctly. The image metrics also seem to be correct (although I haven't tried with super corner cases like ptoing).. So I guess that the bug is about the arrays being indexed incorrectly.

I'm not sure what you mean. They only effect how the number of levels (which can be calculated from a bit depth) are evaluated at 8-bit per channel.

Well, to my mind, bits per channel is a compact way of specifying levels per channel --  2 means 2**2 == 4 levels, 4 means 2**4 == 16 levels, etc. Which is why it made sense to me that the setting for how these N intensities were mapped to the 0..255 continuum would be shared between them -- bpc is just a different notation for entering number of levels.

Yes, the current formatting is an improvement.
The Levels readout to help easily tweaking custom levels is helpful, too.

[surt]

Levels per channel seems to be quite fucked if you do weird numbers, like 2, 3, 255. The output makes no sense at all.
Also you should be a able to have 256 steps not just 255. 0-255 = 256 steps.

Whoops, yes.
Fixed the level count. Though with all 256 and swatch size bigger than 1 will probably produce an image too big for the browser to handle.
I had the green and blue levels mixed up so if they were different it'd be gibberish. Fixed now.
Still something wrong with BGR.

[Ai]

Whoops, yes.
Fixed the level count. Though with all 256 and swatch size bigger than 1 will probably produce an image too big for the browser to handle.
I had the green and blue levels mixed up so if they were different it'd be gibberish. Fixed now.
Still something wrong with BGR.

Just did some more testing. The cases I report above are still bugged, for BPC mode, and Levels per channel still seems to have a bug -- eg levels (2,3,4) == wrong(only 3 levels of B are used, the fourth tile uses #0 again)

EDIT: can reproduce on both firefox and midori.

[ptoing]

rgb443 is also totally fucked. You are doing something weirdly. Actually in the build that is up atm everything but cases where all numbers are the same fuck up in some way. You must have some bug in the way you draw in the values.

Edit: Werks now. Good jerb.

[surt]

Oops, uploaded to wrong directory. Try now.

[Ai]

hmm, yep, all of my test cases look correct now
Oh, btw, just tried Centre intensity mapping, I like the softening effect -- basically the same thing I was trying to achieve with my curves adjustment. I'd probably edit it afterwards so almost-white was actual white and almost-black was actual black, though.

[Helm]

surt went and talked to the Guardian, he told him he'd fix all his code if he represents.

[ptoing]

Surt: why can I not enter just 0 bits or levels as well? What if I want a palette without one of the channels?

[surt]

Surt: why can I not enter just 0 bits or levels as well? What if I want a palette without one of the channels?

Because I forgot that 1-1=0. Not zero levels (must always be at least one level of each component, even if that level is zero), but zero bits now.

Oh, btw, just tried Centre intensity mapping, I like the softening effect -- basically the same thing I was trying to achieve with my curves adjustment. I'd probably edit it afterwards so almost-white was actual white and almost-black was actual black, though.

I've been thinking it might be usefully to add some post processing functions such as gamma/bias, gain, brightness and contrast to allow some colour fiddling? Separate values per channel?

[||||]

I know it's an old thread; but I really wanted to try this exercise again.
 I used a very incompatible image as a reference/inspiration (lots of thin lines and curves; not a very cluster friendly original image).

I strayed a little bit from the original purposefully in some sections but tried to keep the composition close to the same.



Big thanks to Kasumi for helping me seek and destroy 19 single pixels I missed.

The original image by Sci-fi master Frank Kelly Freas can be found here: https://s-media-cache-ak0.pinimg.com/originals/19/87/eb/1987eb92144a86738aa253abf788acf9.jpg On the actual book cover the face is much more blue-tinted also.

(edited: one elusive loner walked the wrong street and got killed)