Fad Avatar

[LoTekK]

Started making a fad avatar of my girlfriend tonight. I more or less used the same pallete as the standard fadatar, though I think I added a single intermediary color somewhere along the ramp. Currently at exactly 16 colors. Here's where I'm at, as well as an anigif of the process so far (I didn't save a copy of the sketching stages).



Crits very much welcome.

edit: Bah, that was a stupid mistake. Links fixed.

[Helm]

Your hosting seems to be broken.

[TrevoriuS]

Meh, the links are just wrong... THey're formated like http://image.gif/

[LoTekK]

Doh. That was completely idiotic. Links are fixed. :p

Also, quick update. shifted the eyes up, among other things. Not sure if I should be dithering the sweater.

[Helm]

Here's an edit



Grab a cup of coffee for BLOCKTEXT APPROACHETH

First of all the good: This piece is pretty wonderful, especially in the color choice. Plus it has character, the person shown has human qualities, isn't just a collection of symbols. This is commendable, I always am very impressed by this ability to give anima to a drawing and I am personally off the mark on this account a lot. That being said your art lacks polish. Polish is what makes an ok piece of art into a great piece of art when it comes to pixels. To achieve this level of polish (well, 'a' level, my edit isn't up to where you should take it, I think, you could go even further) isn't just a matter of smoothing things out willy-nilly or adding more colors (I didn't) or anything like that. It's a matter of conceptualizing space between pixels and using the grid you have, to approximate a level of detail that is smaller than a pixel. To do this you must think about the aspects of pixel art that help you to do this, and avoid the various pitfalls. On one hand, you can use subtle gradation with the tones you have to suggest slopes of the planes that in the mind of the viewer register on a more platonic, ideal level than what the clumsy hugepixels usually achieve. You need to be subtle to do this. Keep in mind that the pixel art artistry might be something we pixel artists zoom in to appreciate for minutes at a time, but for the viewer that doesn't know how it is made the impression is made in the first second. If you've convincingly faked depth and sub-pixel detail, the end effect is of immense important on a layman. So for example, you are missing this point when you let the side of the black hair un-AA'ed, so on.

That's one point, the other - and one which has not been properly documented in pixel art theory so far as far as I can tell - is how stripes of pixels create banding effects and how these banding effects severely damage that fake subpixel immersion effect that we want to go for. Banding doesn't only occur in the blatant forms, virtually every time you line up vertically or horizontally a line of pixels of one color with a line of pixels of another color, you are creating a banding effect. Almost all the time, to the detriment of the piece. It would take a multi-page tutorial to fully explain just how big a deal banding is, and it took me years for my eyes to see it finally, but now I am obsessed with it, in pixel art theory I place it above almost all other effects one should be wary for. As we often say here, the art is all about specific pixel placement and how every pixel relates to the ones around it. Well, banding is an effect that is both specific to each pixel and holistic in the forms they take and as such is elusive but very very important to grab hold of.

Banding in fact, occurs most on the polishing stage. Too many colors is a surefire way to get clumsy banding effects anywhere. When pixel artists graduated on their respective platforms from few-color mastery on the NES or the EGA palette or whatnot to more colors, VGA cards, 16bit engines and so on a horrible thing happened: they thought more color meant better art and just slapped huge amounts of different shades on relatively simple shapes and then their lack of understanding on how to do volumetric lighting on one hand, and how to avoid banding on the other, truly showed. More colors aren't always good, and they never are if you don't know what to do with them. Your palette is robust (a bit muddy on the mid-tones perhaps and a bit scarce on the darker tones, I amended the first issue on this animation but not the second) and the volumetric is pretty sound -- besides the hat. You shaded it inversely which shows a lack of understanding on how that volume would work. Are you working from a photo on the face? Because it's removed, skill-wise from what's going on on the hat. The subtlety required for the AA work and the banding-killing is really difficult to express into words so I would just urge you to take the animation in a program of yours and look at it zoomed in and zoomed out, part by part and actively challenge yourself, ask him why does he think this part is better in the helmversion than in the original one. Be analytic, dissect my choices. Ask me if you don't get why something is as it is, but really from a point and onwards it's about your own observations and not how well I can explain the minutiae of polishing a pixel art piece.

I will give you a hint though: look at how many parts in your piece there are where lines or shapes of color line up vertically or horizontally, and how few such parts are in my version. It's a bit like a strategy game from one point and onward, a bit of give and take on each shape so as to get the least amount of banding without giving away the needed edges and definitions of the shape.

[LoTekK]

Good god, helm, thanks a shit ton for both the writeup and the edit. I really appreciate the compliments, and even more so the crits and advice. It makes me happy that you see not just life in the piece, but an individual, since that's one thing I always try to strive for, especially when doing personal pieces (and in particular, portrait-type pieces). Just a note, though, the pallete isn't actually mine, it's just cribbed from the [standard?] fadatar pallete (with one extra color and possibly one or two hue shifts, though I cant recall if I kept those).

Before I get to responding, I'll put up the current wip and another process anigif. This was done before I noticed your massive post, so there's tons of stuff as yet unaddressed.


Tweaked face shape, hat, a few other things. I think I may have to tweak the face shape more, though. Need sleep for fresh eyes. And I don't know why I brought the shoulders in so far... That'll need fixing tomorrow...

Also, I should probably link to the ref I'm using:
Ref photo

I'm a tad sleepy, so a more in-depth response may have to wait til tomorrow, but I'd like to address a couple of things. Alright, first off, and in my defense, the piece is far from done, and I basically put it up for crits at an early stage to see if there was anything ridiculously wrong with it to date (eg, the hair wasn't completely AA'd yet). That said, your points and edit definitely helped clear up some questions/doubts. For one, I wasn't too sure how much selective dithering I could get away with in order to reduce banding and increase my perceived colour pallete. Your edit definitely opened my eyes some as far as smart dithering is concerned. I think I have to get into the habit of referring to the 1:1 preview in ProMotion more. Too often I tend to focus on the 8:1 working view and think the dithering isn't working (ie, at 8:1 my eyes don't benefit from visual mixing).

[edit]
My point about this being wip was meant as a response to the comment about lack of polish. While I'm under no illusions as to whether I can polish a pixel piece to perfection (still plenty of work and learning to do on that front, among others), I just wanted to point out that I'm still noodling with "blocking", so to speak, trying to nail not only a likeness, but also her personality. That said, I'm still taking your advice on polish to heart, as it was all very useful.
[/edit]

To your point about faking sub-pixel detail, that's probably one of the things I enjoy the most about pixel art, and the nostril and eye areas were probably the bits that taxed my brain/eyes the most in that regard. I still have tons to learn, but at least it's an enjoyable sort of learning.

The hat I actually hadn't gotten to properly at the time of my previous post, so the shading was a tad funky, though it was blocked in based on the ref photo (the right side was kinda botched, though, and I overdid the underlighting on the visor bit). I've addressed that in my latest wip, though it's still as yet unfinished.

Once again, I'll probably reread your massive response after a good night's sleep and respond accordingly (ie, gushing with adoration ), but I wanted to thank you for taking the time out to writing all that.

[edit2]
More wip-ness:


At this point I think I'm noodling a bit too much. Fun mental exercise, though, shifting things a fraction of a pixel.
[/edit2]

[edit3]
Latest wip. It's coming along, and I'm much happier with the likeness now. Still experimenting with the dithering, etc. Not entirely happy with the cap yet.

[/edit3]

[Helm]

This is progressing greatly. But do tell me, what do you think the single-pixel noise you have on outlines and other places is achieving? I don't mean the question in a patronizing way, I really would like to discuss the pros and cons of such a method.

[LoTekK]

Thank you, Helm.

About the noise, to be perfectly honest I guess I'm still feeling my way around this. I guess I'm thinking in terms of impressionist painters, using optical mixing to suggest more shades than are available in the pallete. However, I'll be the first to admit that I may be overdoing this. Also, with some of the shades, the pixels are standing out a bit much in areas (the two oranges are a bit saturated and stand out a tad if I over-use them in the dither). What's your take on the dithering? Too much?

For the outline noise, do you mean the straight line dither (which as of the previous wip was limited to one section of the hair outline), or the way I've tried to AA the curves? If it's the straight line AA, I think I got the idea to try it when I studied your edit (which reminded me of your thread talking about dither smoothing). The AA of the curves is something I'm also still trying to get my head around, though in this case I'm basically buffering the curve and looking at the 1:1 view to make sure things are smoothing out instead of making the curve jagged. Essentially a bit of theory, a bit of trial-and-error. In some areas I think it's working, while in others I'm still of two minds of it (the upper edge of the cap, in particular).

New wip. After stepping back and studying the proportions, etc, I decided to make some more changes. Made the face a bit narrower, and shifted her right eye half a pixel over to our right (that was a bit of a mindjob).

[Helm]

Thank you for being lucid in discussing your art choices, it's not very common.

The thing is, you are taking an outline of a form (let's look at the top of the cap) and you're peppering (I think this term is a good one "peppering". It looks like pepper!) it with darker pixels on its insides. There are two questions which arise from this:

1. Do you need an outline at all there anyway? Wouldn't it just be  body of color (the cap) being hit by a lightsource (the sun) against a bright background (the yellow)? Wouldn't that situation 'outline' so to speak the forms against each other adequately without going and adding a darker lineart on top? Don't think so much like a spriter when it comes to outlines. I can't tell you to outright forget about them, they're useful, but not like this. Because

2. You are sabotaging your lines with the 'pepper'. It comes out a bit like jpg artifacts. If you're going to keep an outline (darker or brighter that the body of color it is surrounding it doesn't matter) and you want to vary its values as it goes (which is a commendable sport) then I suggest you wait for a 'step' (think of a ladder) before you change the color. Let me throw up a quick example.



Give your linework space to breathe. A single pixel is a bit of a terror. Alone on a flat field its an amazing specular highlight that no other art form can do as opaquely. but it should be used in moderation. Single pixels do not rule your piece, they emphasize the effects that bodies of pixels convey. Look at your current piece and think to yourself "what is the ratio of single, visible pixels, to the flatter surfaces?". If it is more than 30% or so, I think you'd need to tone it down. Faking color mixes is all well and good, but not to the expense of the power of a flat color.

[LoTekK]

Actually, while I think the lucidity comes partly from the nature of the medium, a lot of it comes from being able to actually discuss the process with someone knowledgeable (so I should be thanking you ). While typing out my response, I realised that I had taken the "smart, selective dithering" that I had pointed out in your edit and went all overzealous with it, in the process forgetting the importance of nicely-defined planes and points of interest.

On the outlines, I had actually been struggling with that, wondering if I should just get rid of the outlines in some of the lighter areas (cap, possibly shoulders). I'll probably go ahead and get rid of some of the outlines on the cap, at the very least, but thanks much for the diagram. It definitely helps to clear up some of the issues I was having with the outline. And yes, peppering seems a more than appropriate term to use here.

Time for bed, but I'm looking forward to tackling this tomorrow with a fresh and rested pair of eyes.