[Davem]

Now I've textured a few game characters from time to time using different methods, but i wanted to find out how everyone else tends to do it. In the past I've tended to do it mainly by physically breaking the mesh apart and laying it out flat, planar mapping, then morphing it back to its original shape.

I've also done it by using selected faces (or material ID's) mapping those faces independantly and laying it all out in Unwrap UVW (3DS Max). The main benefit of this method was that it made no changes to the actual topology of the mesh and so the mesh can be being boned and animated by another team member before or while the textures are made. That way the meshes stay identical and UV's can be loaded onto the animated / boned mesh. Does anyone find that this usually occurs in production or do ppl tend to model and texture before it gets passed to an animator etc?

A good technique i've used before is to use a coloured grid map when laying out the UV's to make sure that the texture stays in relative proportion to its face sizes.

Is anyone out there using Deep Paint 3D or even Chilliskinner?

I've also noted that some teams try to keep the different parts of a character mesh inside power of 2 areas on the map whereas other tend to just lay them out arbitrarily. An example of the power of two thing i saw on: joachimart.com

Anyone got any comments?

Cheers

Dave
[email protected]

[Brain]

I use the second described way; selecting faces, assigning an ID, then applying a checker pattern and correcting any distortion, then exporting with Texporter, painting it up and voila @:-)

[Sumaleth]

UV-mapping is the most AWFUL job in the whole process of model making. No doubt.

I UV-mapped several dozen polygonal models last year (1200-5000 polygon range) and ended up using Maya for the process;

. I made up a 'checker' texture 1024x1024 in size (4x4 pixel checker pattern) that I applied to the model.

. I'd break up the model into all the separate pieces I wanted for the texture layout. (and assign differently tinted versions of the checker texture to each piece)

. Have one portion visible at a time; project some starting coordinates using whichever method is best for each piece (planar, cylindrical, spherical).

. Then move each UV vert separately until each polygon had the least amount of distortion possible and the checker pattern was the same size across the object.

. Once all objects are done I merge them back together and resize all the UV-meshes separately until the checker pattern is the same size across the whole model.

. Then resize all UV-coords so that they'll fit into the 0->1 unit box.

. And finally position/rotate each piece for optimal skin layout.

--

An ugly process, but the skins I ended up with were as good as you could get, and certainly a LOT better than any of the 'automatic' programs i tried.

We had an in-house util that took UV coords from my version of the model and put them on the animators version of the model - this was needed because by the time I'd finished with it the polygon arrangement was completely different.

Oh, the reason people tend to use textures that are factors of 2 in size is because that is the way 3D accelerator cards work. They only do sizes like 256x256 or 512x1024 or whatever. If you have a texture the size of 100x400, for example, then different cards will treat it differently, but the end result is the loss or blurring of details, so it's best to make the texture at the exact size the cards will use.

Row.

[Davem]

About the power of two thing i wasn't actually clear i think. I know the entire map has to be a power of two but what i was referring to was internally dividing that map up into power of two areas.

For example the face of a character in a 256 x 256 map would sit inside the 128 x 128 area of the map. From what i understand its meant to be more efficient for certain hardware to draw.

Anyone doing this type of thing?

Dave