I think I have to say basically the strict category is it for me too. As we go left or right from there it seems like uh 'senseless technique-porn' (that is, considered as pixel art, it seems to be posturing 'look how linear I am!' / 'look how smooth I am!', rather than just solving the pictorial problem in question.)

It should be noted for context that I am positive towards the use of pretty much any tool (definitely including indexpainting and [blurs that are more intelligent than Gaussian is]). But another way I can put the above paragraph is: I think the tool is controlling the user, more than the user controlling the tool, in those broader items.  Perhaps I philosophically differ in this regard, but I think the artist should not 'push around marks' or let the media dictate to them; The influence of media should show in the 5% of the picture appearance, not the 95% (which should be showing the things that need to be shown, in correct relation/priority).

An interesting thought I had regarding this was about gradients - basically when does a "blur" or gradient turn into a "ramp" of colors which we consider pxelart like. I think the "pixelart feel" also has a lot to do with a certain amount of colors on a certain level of size.
Like something with higher reslution can use more colors to look like pixelart, while if you just blow up a small sprite with the same colors it can, but not necessarily has to feel "not shaded enough" and not as much as pixelart as the smaller version with the same amount of colors.

IMO that falls into the same 'meaningful decisions' theme that some people like to use when talking about this. That is, a gradient just looks like a gradient, if each shade doesn't have 'individuality'. When each shade has an individual shape that feels like it is defining particular subtleties of form, rather than just 'snowballing' with the other shades in the meta-cluster to create a gradient, it's a ramp. When you have too many shades, it just feels like you are nesting them, rather than 'pushing' one shade against another to exaggerate the effect of both.

It seems to me straightforward that the amount of individuality a cluster of pixels can have, is limited by the number of pixels comprising the cluster. So we need less clusters + more .. 'forceful' (gestural/heavily articulated) clusters to achieve the same 'pixel art feel' in smaller canvas sizes.  Conversely, when we blow up a small sprite with say HQ3x, it feels coarse mainly because the clusters aren't really utilizing the extra articulation that's available, just 'bevelling off the edges'. It has to stay 'finely fitted' to the particular canvas resolution, or else not feel like pixel art.

IMO the possibility of using dithering in larger sprites while it generally just looks ugly on smaller sprites, is a consequence of that same principle.  Dither can participate coherently in metaclusters on large PA, but tends to just confuse the identities of all neighbors in smaller PA.

Some techniques seen in old Amiga demoscene pics (particularly low color ones) are also working along same lines - making intentionally elaborate clusters to reinforce this feel. This works well IMO (reinforces 'pixel art feel') but I don't think the other aesthetics resulting from it are very good (excessive detail/eyecatches -> compositional problems)

Yo, don't draw every day.
It's an epic bold statement that sounds good.
But easily leads to RSI in the hands, a decline in mental health, and lacks life balance.

IMO a strong and steady pace *does* need to be every day (with the exact *amount* per day being more flexible).

15min a day is IMO a good way to start out, to develop the drawing habit, and I would say it is in no way excessive or unhealthy[1].

I'd suggest 15min as a minimum amount because it takes a certain amount of focused effort before your feelings start getting onboard with the idea of 'working on this drawing task for no less than 15min'. Sessions shorter than this are IMO not effective at reinforcing that habit of focus, which is absolutely needed to learn effectively. [2]

There are also natural opportunities to draw in some cases (for example, on buses, or when making notes). Taking this 'lifestyle' approach - don't just give it a timeslot, but find ways to fit it into the context of other tasks - IMO is necessary to ensure consistent progress. This eventually makes the plan of 'draw something every day' quite easy to achieve.


[1] re: RSI, As a side issue: another thing a newbie should be aware of that can have a big impact, both on drawing quality and RSI: Don't press so hard. Using the minimum possible pressure wins on three fronts: less pencil wear, less body wear (ie. protecting yourself from RSI), smoother lines. People often cite control as a reason why they press hard, but the same level of control can be achieved by training your muscle control instead (particularly, working to draw mainly via shoulder motion rather than  elbow or wrist motion).

[2] Learnt this from Nicolaides' "Natural Way to Draw", FWIW. I also recommend this book on the basis of Nicolaides' ability to teach techniques that *make you think about the subject*, rather than providing a neat way to reproduce the surface appearance.

Thanks so much! The advice got me quite excited so I had to go out and try it. Drawing from life as I'm not quite accustomed to using sources yet.
[...]
The above is, admittedly, a tree that is easier to draw. But your advice about the spaces between the trees ballooning from a central point helped with this one. I still need to figure out how to articulate particular leaves it seems. The L-system is really interesting, checked it out and the concept definitely changes the way one looks at natural forms.

I don't think there's any per-day restriction. Personally I just try to think about how much vertical space I'm taking up, and put some images in spoilers to reduce that if needed.

Glad it helped. Right now, I need to go to bed, but some quick comments:

* Nice consistent tone application, you're obviously comfortable with it.
* Not such a fan of the zigzag motion on foliage, it directly creates tangents.
* Pretty respectable planes
* Except at the base of the tree. which feels like it isn't turned in 3d near the lowest joint.
* Left Hand of Darkness sounds like it might be funny, but nope, not the reference.
* Don't know what is going on with the line moving towards top-left. Is it supposed to be the edge of a block of leaves ?
* Yeah, ground plane was added digitally. It really helps a lot IME, I'm trying to get in the habit of putting a ground plane clearly in most drawings to begin with.

Tones (digital part is intentionally minimal)



You can probably tell which one came first.

Quote from: Ai on April 17, 2017, 10:47:30 am

I like the looseness that comes from indexpainting with a truecolor 'backbuffer' (this is basically how it works in GIMP 2.9) -- you can 'invisibly build up' paint until it passes a threshold and becomes visible.

Please explain. How do you do this in GIMP?

It's actually pretty easy (assuming compilation isn't scary to you

* Compile and install a recent version of GIMP 2.9. Ideally the very most recent. Installing deps (especially babl and gegl) may be required.
* Launch GIMP and create a new image
* Prepare a palette with the colors you wanna start with.
* Convert the image to indexed with custom palette set to the aforementioned palette. Uncheck the 'remove unused colors' checkbox.
* Paint with anything that isn't pencil tool. Airbrush, or paintbrush with a fuzzy brush, for example.

Experiment with the new Force parameter to get paint coming quicker or slower. Everything is fully functional (eg. use whatever paint mode you want, it'll work)

* Also make sure your painting is using Default, not Legacy, blending (that's controlled by the new widget sitting next to Paint Mode in the
tool options). This isn't strictly required but makes everything much easier to predict.

* Layer modes work as expected, except the result of blending is not indexized before display. IMO this is a bug. Need to merge layer down to fully indexize. Non-binary alpha and masks also work as expected -- eg. I used Levels on alpha channel of text layers in images below, to get a likeable indexized result.

Basically you can use any tool as you would on an RGB image; the 'pushing through' is really an effect that comes about due to the fact that you're painting in RGB but the displayed surface is quantized to the given palette, so the change only shows up when the "real color" is different enough to be quantized differently

Some indexpainted demos of Force vs Opacity (using standard brush tool w/ white fuzzy brush); hopefully this illustrates that Force works similarly to [Flow in Photoshop]

(Force)
(Opacity)

If this is not an adequate explanation, I was thinking about doing a video. Let me know if you'd be interested.

EDIT: actual sketch content: Quick test of Mypaint's new snowflake symmetry: