Cyangmou: Well, First of all, I'm tickled you took a look at it, at all! Thanks.
And I completely agree, this is in no way pixelart. It's a 3d model with trickery - but if I can make the tricker fool the eye into thinking, even a little, that it is pixelart, then I will consider the experiment and method to be a success! I can get the outlines cleaner, and reduce the colors per plane, as you said - I've made some major improvements on those fronts here lately. As for the model pre-processing to get the right input into the shaders, it's really not that much of an issue - most is done automatically. The rest is just building the model with the results in mind - literally no different from any other type of 3d modeling. I totally agree that the mech isn't the best shape for this sort of thing - that's entirely intentional. If I had started with a shape that was easy to translate into pixelart, then what's the point? This is all about doing things the hard way to make sure they work.
Ai: Yes - there's no total agreement what pixel art is. Which actually works in my favor a bit - I can definitely edge closer to an indeterminate goal than I can to a hardline set of exact specifications.
ptoing: Yes! I can do that exact thing, and have - it's in the screenshots below and in the build! It actually is a great addition - non-shadowed models now have a ton more depth than they used to. One minor downside is that angled view of objects can look a bit odd - see the fighter or 3/4s view mechs. Because it's shading the back half of the model from the camera's view, it's not really taking the lighting into account as shadows actually do.
Gil: and yes, that's exactly what I've done. It simply changes how the lighting UVs are calculated, same as the AO does. I'm also looking into clamping values - but I gotta say, keeping them unclamped allows for a lot more lighting opportunities.
Okay, the build is updated! 3 major additions and changes to speak on:
Pixel Locking: This adds the shader clamping ability of locking each vert of the model to a pixel location - based on the grid of pixels in the camera render. It can be turned on and off, and has the benefit of making some shapes much, much cleaner - while losing some edge detail because of the need to lock to the nearest pixel.
Depth Cueing: The models are now darkened on local object ZDepth - this currently only works correctly on the static, unanimating models - the walking mechs, as they are made up of many smaller objects, don't have the proper information to determine their depth cues. Currently it's a general solution that's a bit too intense in shadowed mode, and not intense enough in Non-Shadowed mode. Working on adding the switch to flip between those options.
Better Palette managing, and AA quality: I've changed the way the textures/lighting ramps work on the objects - they now have a minimized number of colors with a ton of inter-ramp borrowing. For example the white paint colors actually use the dark metal colors half way down. This is very similar to how actual pixel art color palettes are managed. This has the benefit of making color clusters more unified, making AA in general use less colors, and forcing the end result to be cleaner. I took the number of unique colors from 160 to 90. I am going to get that even farther down.
I've also sharpened up the AA through what else - trickery - to help try and clean up some of the lines.
Anyhow, on to the imagery!
Please, any further feedback would be welcome!
Enjoy!
