[Greltam]

I was wondering if anyone had experience with procedurally generated sprites?
I'm in the planning/feasibility part of a new idea I want to work with. If anything, It'd be like Spore but 2d(I haven't played spore but I can across it while researching.) I also came across a site that's been semi inspiration. Pixel Robots details procedural generation, which also spun off to a Pixel Spaceship generator.

My idea is to have a String of set size for the genetics, and generate a sprite based on that. I don't want to constrain the sprite to fixed size, so that some creatures will be small, others large. I would also like to have the sprite "grow" from it's birth(say a few pixels) into its full-sized form.

Idk.. it might be something I have to do myself. At this point I've got just random pixels sans color.

[Greltam]

Noone probably cares, but I found a method to at least procedurally generate plants. L-System's can produce branching structures that resemble plants.
If you flip it upside down it looks like a bush.

[stackoverflow]

I suppose you could also use a short random walk, then shorter random walks from random portions of the first random walk. I think that would produce a similar looking bush?

[ButchDean]

The trouble with these procedures is that unless they are properly implemented you will end up with unexpected results. I really don't know where you are going with this, but I suggest you leave it alone for now and get a game developed without procedurally generated sprites - you're just adding complexity to an already fairly complicated task.

[Greltam]

I think I'll do as PixelRobots/PixelSpaceShips does, and predefine a few sizes/areas within to denote head/body/legs/arms for the animals. I'm thinking of sticking with L-Systems for the plants. I wrote a prototype for plant growth, as I wanted to see if it would work. Highly unoptimized.

"Plants" growing

[ButchDean]

Very nice work, Greltam! I feel you need to apply yourself to 3D work now. I feel you have the drive to give it a good go.

[stackoverflow]

That's cool~ I can imagine those in a game. How hefty is the algorithm on resources?

[Greltam]

@StackOverflow It was bottoming out my CPU with just 10... I reworked it alittle bit, and now It runs a bit better. I have a list of the shapes that constitute the "plant", lines and ellipses. These are then pulled out and put into their respective list of branches or leaves. Here, for plant growth, I have two different methods. Redraw the whole plant to keep the branches behind the leaves, or draw the branch/leaf over everything.

2.0GHZ Dual-Core CPU
With 100 plants onscreen

at 30 updates(plants growing) + 30 screen draws:
sequential shape drawing/no screen Draw - avg Cpu: 8% avg RAM: 100MB
sequential shape drawing/screen Draw - avg Cpu: 65% avg RAM: 100MB
redraw whole plant/no screen Draw - avg Cpu: 65% avg RAM: 400MB
redraw whole plant/screen Draw - avg Cpu: 100% avg RAM: 400MB

at 10 updates + 10 screen Draws:
sequential shape drawing/no screen Draw - avg Cpu: 3-5% avg RAM: 100MB
sequential shape drawing/screen Draw - avg Cpu: 25% avg RAM: 100MB
redraw whole plant/no screen Draw - avg Cpu: 35% avg RAM: 200-350MB
redraw whole plant/screen Draw - avg Cpu: 55% avg RAM: 200-350MB

10 updates/"plant growth" per second is also pretty fast. I think 3 per second would be alright. I'm also using Java, and could be shuttling the images to the graphics card so screen draws wouldn't incur as much on the CPU.