Kikito is right. I don't think there is much need to do very detailed models, most of it would get lost when you scaled things down. The trick is to know what kind of angle and distortion you are viewing your world in. You definitely don't want to take a screenshot of the model while viewing it with perspective. And you might want to scale things so that a unit of width is the same size as a unit of height. Other than that, I don't think there is anything special to modeling for pixel art.
Some wireframes:
1rcxm.png (47.91 KiB) Viewed 1290 times
hAUOr.png (40.48 KiB) Viewed 1290 times
A smoothed treestump:
K3eWU.png (95.85 KiB) Viewed 1290 times
Scaled down:
GAj0H.png (2.8 KiB) Viewed 1290 times
So once I'm viewing something like the smoothed image of the treestump, I literally take a screenshot, crop out everything but the stump, and scale it down using nearest neighbor. That's it.
Of course, you have to program the shaders as done before.
GarbagePillow wrote:Kikito is right. I don't think there is much need to do very detailed models, most of it would get lost when you scaled things down. The trick is to know what kind of angle and distortion you are viewing your world in. You definitely don't want to take a screenshot of the model while viewing it with perspective. And you might want to scale things so that a unit of width is the same size as a unit of height. Other than that, I don't think there is anything special to modeling for pixel art.
Some wireframes:
A smoothed treestump:
Scaled down:
So once I'm viewing something like the smoothed image of the treestump, I literally take a screenshot, crop out everything but the stump, and scale it down using nearest neighbor. That's it.
Well that's weird. Thanks for attaching it Nixola.
That is a good technique. It is essentially bump mapping. Another way to do it would be to generate the normals from the bump map with a pixel shader. http://http.developer.nvidia.com/CgTuto ... ter08.html explains how to do that.
I don't think the end results will be as accurate as using 3d models, but if you already have sprites and just need normals, this would be a good approach.
GarbagePillow wrote:
I don't think the end results will be as accurate as using 3d models, but if you already have sprites and just need normals, this would be a good approach.
Thus, it could be easier to be done for people that can't / doesn't know 3d modeling (me ), or for a large amount of tiles and sprites.
), or for a large amount of tiles and sprites.
