Aha that explains the sudden reduction down to two shades of green in the background. I hit the no dithering button (which you would to keep the size down). Didnt notice the green was dithered, I just thought the tool had done some very weird colour quantisation. So if the original image had spread the colours more evenly amongst the 32 it had (say 6 shades of green), it would not have looked so bad at 3k.
One of the problems I had in my old demogroup was the artists kept complaining about my code as they never got to be involved in small productions (you cant squeeze much art into 4k). On the other hand none of them was interested in size art: doing something beautiful in very few bytes. I even offered to write some tools but nobody was up for it. When I asked for creative use of 2 colour bit maps for a horror 4k, I got nothing: and there were a lot of artists in the group.
So in my defence :-) I'd say I'm often forced to butcher high-res "art" or avoid it completely until the glorious day I find an artist who wants to work in 500 bytes.
To be fair its the same for coders, we can do a *hell* of a lot more in 30k than in 4k but the competition is for 4k.
What would be interesting I think is to look into some sort of tool for size art. For example, shockwave, if I gave you a tool that could render a shaded 3d ellipsoid as a single command...then that green background could be done in (x,y,colour,xr,yr) or about 10 bytes but it would be rendered in 32 bit colour. Not having any artistic skill, I have no idea what would be useful and what not. On the flip side, I do know what could be done in few bytes in OpenGL.
Hmmmm...thinking out loud here, rambling, off topic. Sorry.
