next »

[DanFessler]

I'd just like to give my 2 cents as a professional pixel artist.  The term "pixel art ruleset" has been thrown around and I cringe every single time.  There is no "rule-set" to pixel art.  You can create your own, for consistency sake, but that'd define a pixel art "style" more so than an overall rule-set for pixelart in general.

with that said, there are definitely a few things that goes into all pixel art...

Limited palette:
When making pixel art, the goal is to describe the most out of the absolute least amount of colors.  Lots of work goes into making a really great palette.  A versatile palette can reuse color indexes in clever ways.  Instead of having a green ramp and a blue ramp, a good pixel artist will share colors across both ramps.

If you color-reduce a 3d-rendered image into a pre-defined palette you'll find an undesirable effect at times.  This is because pixel artists will simplify the planes of the object to work well with the limited palette he has to describe it.  "cell shaders" achieve this same effect with 3d rendering.  Some artists skirt around this by applying dithering to simulate more colors than he has at his disposal.  While this works, its really a preference to the artist and usually he still mindfully describes the planes before dithering anyway.

Dithering:
Dithering simulates a gradient of colors by the alternation of two colors in various patterns.  Dithering can be random, or structured.  Most pixel artists only use structured/patterned dithering.  Usually dither patterns are all the possible alternating patterns within an 8x8 square.  And even then, they usually limit their dithering to 25%, 50%, and 75%.  When applying dithering to an image, pixel artists will often be selective where they decide to apply it.  In terms of applying this to a 3d render - this would mean there'd need to be some sort of threshold to determine when/where to apply dithering to an image - making sure if the gradient is too steep, it wouldn't be applied.

Anti-aliasing:
nearly any image utilizes anti-aliasing.  The difference here is the AA is applied within the pre-defined palette.  Mathematically, you're using the average color of two or more shapes that go through a square (or pixel in our case) - in pixel art, you then match that average to the closest color in your pre-defined palette.

Outlining:
Much of pixel art uses outlines.  This is a style choice, but if you chose to apply it, there are some things you can do to make sure it is pixel art friendly.  I've found a quick way to apply an outline to a sprite automatically is to always outline with a color that is one or two shades darker than the thing it's surrounding.  I'm sure there are rendering filters that could generate an outline effect on a model.  You'd just need to make sure that the outline color is dynamic and stays within the pre-defined palette.  if you cannot find a suitable outline filter, you could fake it by duplicating the model, inverting it, then expanding it a bit.  Then you'd need to post-process the outline color to do as I described earlier.

anyway.. that's my two cents

[Tycho Brahe]

I was writing a post about the ways to address those issues a couple of days ago, but i gave up.

basically, most of the main features you've described can be fixed with some simple algorithms, all it takes is applying them to the image together. For example, there are fast edge-finding solutions out there, and all it should take would be to apply that to an image, whilst it was being scaled down.

Of course, its never going to be 100% perfect, but by applying all the techniques together it may be somewhere close.

[Alex May]

Well for outlines we have 3D algorithms that work on 3D models that can do this. So that could be covered quite easily.

The things that I think are missing are large amounts of detail, which it's possible to create but would take a lot of effort, and really good lighting, which a lot of pixel art really depends on IMO. Really good lighting is also something which 3D graphics is only just starting to be able to do.

[Tycho Brahe]

Really good lighting is also something which 3D graphics is only just starting to be able to do.

What?
3d engines can support multiple differently coloured lights, dynamic shadows and a whole load of other shit. I think what you mean is that 3d engines cant do STYLISED lighting well, pixel art often uses slightly over the top, non realistic lighting, which gives it its character, 3d engines by comparison aim to provide highly realistic, detailed lighting to make the game as close to life as possible.

[Nektonico]

Really good lighting is also something which 3D graphics is only just starting to be able to do.

What?
3d engines can support multiple differently coloured lights, dynamic shadows and a whole load of other shit. I think what you mean is that 3d engines cant do STYLISED lighting well, pixel art often uses slightly over the top, non realistic lighting, which gives it its character, 3d engines by comparison aim to provide highly realistic, detailed lighting to make the game as close to life as possible.

Again, its a bit ambitious to want to do this ingame when we still cant do it for prerendered sprites (in a 3d modelling app of your choice). If one managed to do this efficiently you could do away with pixel pushing (to an extent), and just prerender 3d shapes to a pixel art style (doing isometric stuff or drawing/pixelling a complex object from multiple vantage points and mantaining consistency (and animating it) is a cumbersome process).

Maybe others have attemped to do this in the past but pixel artist hitmen paid them a visit to keep the technology from the public. ... Artistic luddite conspiracy i say 

[gimymblert]

I'm not seeing a lot of game taking adventage of subsurface scattering enough

[Rob Lach]

You definitely won't be able to achieve the effect you want by filtering 2d images.

I don't know exactly how it would work, but I think if you have a 3d model where area is colored as a solid color, and then used the lightining calculation to apply a dithered style of self shading, then add on some edge detection you might be able to get there.

[Alex May]

Really good lighting is also something which 3D graphics is only just starting to be able to do.

What?
3d engines can support multiple differently coloured lights, dynamic shadows and a whole load of other shit. I think what you mean is that 3d engines cant do STYLISED lighting well, pixel art often uses slightly over the top, non realistic lighting, which gives it its character, 3d engines by comparison aim to provide highly realistic, detailed lighting to make the game as close to life as possible.

I'm talking about reflection, refraction, radiosity, etc etc

[slembcke]

Wait? Pixel art uses any of those?

[Alex May]

Can do. Depends on the art. The picture in the OP certainly isn't a low res detail-free pixel creation, it's really detailed, shows complicated and hard-to-simulate (i.e. artistic license) lighting as well as soft shadows (like how ambient occlusion works) and hard shadows, a degree of HDR...

And that's just the aesthetics. Getting the pixels right is the other part, as simply rendering the 3D scene out will come with inaccuracies that won't look like pixel art, and let's not even get into how you'd go about drawing the characters at that resolution.

Then we're asked to be able to zoom right in to see the grain of the woodwork or presumably similar detail in any other part of the image.

You could probably work a small way towards this as an academic department studying NPR, over the course of several PhDs.