Error Diffusion Dithering

Hey I'd love to see if anyone can figure out this with me. I'm looking to make a filter that will convert an image to a pixel dithered output. Something like this:



Or like this with multiple colors:



I am specifically looking to make error diffusion or false error diffusion like Floyd-Steinberg, and not ordered diffusion. Ordered diffusion basically inputs a small grayscale bitmap as the diffusion pattern, or a diffusion matrix in code that does the same thing.

I'm already trying to get AI to program this for me, with some minor success. It turns out the fact that Filter Forge uses a sample based architecture means it fundamentally misses the required functionality to render an error diffusion.

But, I'm sure there's some novel ideas out there. I don't know how to code any of this myself, and I am only about as useful with LUA as it comes to catching easy bugs or fixing simple structural issues. I did have the idea of having the script build a pseudo grid that will act as pixels, and then have the formulas interact with that grid, but it might be a pipe dream.

I'd love to see what you all can come up with. There's a web app out there that does this BTW!
Dither Me This