That’s super weird. Is there a justification?

A bit rich coming from nintendo, since they do everything to stop people from enjoying their old games.

That’s looking really nice.

I see you have the helmets in separate layers. I have been experimenting with segmenting images into areas of similar colors in order to map them closer to the target palette.

I never got very far, because it was a lot of work. But I think it has some potential.

Also a ton of gray pixels in the right visor. Technically correct, since the original image isn’t quite as saturated as the palette, but it looks weird to me.

The 16 color cga palette is the same as the ega. I just called it ”ega” to avoid confusion with the 4 color cga palettes.

Your version isn’t too bad, but you can see some stray pixels of gray in the left visor.

Normal ordered dithering will use mixes of colors that makes no sense. Like a solid blue area with a handful of bright red pixels. Sure, that is the mathematically correct dithering, but it looks like crap.

This metgod gives *much* better results

I created a palette of the basic 16 EGA colors, plus 50/50 mixes, totalling 73 colors. The key is to use pairs of colors that a human would pick.

Convert to this palette, without dithering. Then clean up stray pixels etc, before converting to the final 16 color EGA palette with pattern dithering.

What are the differences?

Normal. A game is never ”done”. You just have to draw the line at some point and stop working on it.

Oh, you already got it running on macos! That’s amazing. Do you have a binary?

Looks real nifty. Would you mind sharing the project on github? I’d love to try it on Bubble Bobble.

Did you port the existing python/vala or did you have to rewrite?