Photosourcing and UVs
polycounter lvl 18
It used to be that we made our character model, unwrapped it, then textured it.
These days many games make a lot of use of photosourcing, so I wonder if others find that now you end up making the textures before doing the unwrap for characters? That way you can build the head/face/clothes from the photos then distribute the mapping around that.
After thinking about it for a while, I realised that environment artist have been doing this for years - you make the textures from your photos, then map the environment to the textures.
These days many games make a lot of use of photosourcing, so I wonder if others find that now you end up making the textures before doing the unwrap for characters? That way you can build the head/face/clothes from the photos then distribute the mapping around that.
After thinking about it for a while, I realised that environment artist have been doing this for years - you make the textures from your photos, then map the environment to the textures.
Replies
I personally found that quite disturbing because it was quite annoying to fumble around with the the different parts of the texture (esspecially because that was the first time I did it) to make it fit the UVs but I soon got the grip of it, thanks to the seniors that had a very good routine with this workflow.
I also thing unwrapping before hand gives you more spacing control esspecially with model that have a higher polycount.
texturing beforehand is very unprecise and you can't really see the effect of what you are doing streight away on the character wich in my eyes is a huge disatvantage and probably causes more worktime than it saves.
I've already unwrapped enough of my models to know where the UVs are going to go, and using seperate texture sheets for certain parts of the model helps.
Areas that I find it helps to to the textures first is the head. Take a front, side and 3/4 shot, line them up in photoshop with no stretching and squasing, paint them together, then do the UVs aroudn that.
I suspect I spend most of my time doing a quick unwrap, then the texturing, then remapping the model to the texture.
Heres an example of that from the last high-poly head i did(sized down to 25%):
We take photos in advance, building up our morgue, since we know we'll need certain surfaces. But for massaging and adjusting to the model, we need the layout set up and ready. Like PRS-Phil says, seeing them on the model is essential.
Rick your head process seems unweildy since I would think you would run into issues where the shape of the head model and UV-stretching would dictate how you need to adjust the layout of the textures. Like are you mirroring, how are you handling the top of the head, where is the distortion more acceptable?
But I can see your method helping if you need interchangeable textures, like for a sports title where the faces/hands are different. Or for a Sim-style game.
I've actually played around with some rather crazy techniques for doing photosourcing. Here's one that I used for some exterior building models (Max centric):
1. Get flat on front/side/back shot photos of the building.
2. Take into Photoshop, cleanup, eliminate distortion. If desired, add the other building sides/top of the building to make one big decal sheet.
3. Camera Map this image onto a box. This basically projects the image onto the box, but the UVs do not stick.
4. Edit the geometry, add in detail, etc., just putting geometry wherever you need it, so that it lines up with the projection/camera map. Geometry will appear to float under the UVs as you make changes. It's a bit weird, but allows for excellent and fool-proof accuracy in the model.
5. Once you've got it nailed, collapse your stack so that the UVs stick.
6. Unwrap the edges of the little protrusions, such as the sides of window ledges, etc. So that they're not getting streaked UVs that were camera/planar mapped from the front.
7. Continue on witht the other sides of the building, working in a similar manner. Attach/detach elements as necessary.
It is a bit of a strange workflow, but has numerous advantages. First of all, you know that the proportions of the building are accurate to real life. If you are working on a photorealistic sim type game, this is good. Second, you get a very uniform pixel density. Thirdly, the flat-on textures as opposed to a decal sheet with lots of little UV chunks makes for very easy LOD work. There will be little disparity visually between your high and low LODs, with the exception of the silhouette.
Obviously, the limitations this technique are the quality and availability of photos for the subject matter. Still, you can always do some Photoshop fixes and paint over work beforehand.
I wonder if a combo of the two techniques might be cool, like if you used your method as a macro texture and combined it with a set of micro-detail tiled textures. So when you get up close it doesn't turn into filtered goo.
Kevin Bjorke mentioned seeing something like this. Sounds like a smart approach, especially where video memory is tight.