Interesting. I'm curious how much load time we're talking about here... with a semi-complex procedural made of several controls, it can take quite a bit of processor time to render a 1024 bitmap. Even kkreiger, with its not-so-complex textures, took quite a bit of load time to render them all out.
I do see the point though of decreasing a game's transmission size and needing less offline storage.
I don't think that wood's on par with HL2 wood though. Also the rendered shots they show are pretty generic looking, only the bathroom shows anything approaching a more-complex game room (and that's pre-rendered anyhow).
Ahhh... one of the developers explained that they render on the GPU, then compress to DXT and pipe that into VRAM. So... faster than HDD>VRAM.
Anyhow. Generating quality textures using only procedural functions can be difficult. You generally can't get specific shapes other than pure geometric ones (circles, lines, etc.)... so for example you couldn't make Mayan pictographs procedurally.
But some good things are possible.
This was done in a competing product called Filter Factory. They don't use the GPU though.
Neat to see where this is going. Thanks for the post.
So you're really not actually saving anythign with texture memory here tho, this is only useful if you need a game that has a really small installer...... Otherwise you're sacrfaicing quality for no performance gain.
There's no performance gain...yet. I think the biggest advantage of this and why it was used with that robot game is because the game itself is so small, XBL games have to be < 50 megs so you can see where that helped. However I know the guys at allegorithmic mentioned the possibility of some of this stuff being implemented into graphics cards, I think then you could see some performance increase.
I definatly think this is one of those things that could become a useful tool, It irks me when artists brush stuff off like this calling it 'programmer art'. Non-realtime 3d artists use this stuff all the time as a tool and I think that game artists would benifit from it as well.
Replies
I do see the point though of decreasing a game's transmission size and needing less offline storage.
I don't think that wood's on par with HL2 wood though. Also the rendered shots they show are pretty generic looking, only the bathroom shows anything approaching a more-complex game room (and that's pre-rendered anyhow).
Anyhow. Generating quality textures using only procedural functions can be difficult. You generally can't get specific shapes other than pure geometric ones (circles, lines, etc.)... so for example you couldn't make Mayan pictographs procedurally.
But some good things are possible.
This was done in a competing product called Filter Factory. They don't use the GPU though.
Neat to see where this is going. Thanks for the post.
I definatly think this is one of those things that could become a useful tool, It irks me when artists brush stuff off like this calling it 'programmer art'. Non-realtime 3d artists use this stuff all the time as a tool and I think that game artists would benifit from it as well.
Also could be a fringe benefit of upsampling easily for cards with more mem.