I’m currently recording motion capture and as faceshift is using a library of blendshapes, your 3D-character needs to have the same set.
So I decided to use my own face as reference and here it is! My own face in blendshapes and “phonemes” (google it, it’s really interesting!)
Continuing with recording! Also decided on what the face will say in its performance!
If you’re baking you’re essentially measuring the difference in detail between a highpolly and a lowpolly-mesh. Bake that to a map and apply that map to the lowpolly in the render stage,
So! When studios render high-detail surfaces in pre-rendered CG then use a combination of displacement maps(geometry, actual deformation) + normal maps(fake) or displacement maps + bump maps. So I thought this would be a fairly straight forward process. It was not.
The first image is a pure displacement map, lots of nice detail but no fine noise, like very small wrinkles and pores. That’s where the normal/bump map is supposed to kick in to bring that last really fine-detail. But here you’ll get problems.
In the second image you can clearly see how a normal map + displacement map that is baked from the same high and lowpolly looks exaggerated.
A solution to this could be to bake the normal map to a higher geo-level, a “semi lowpolly”. But if you have a bake-engine like Zbrush it will smooth your uv’s. thus giving a mirraging effect like in the third image.
How do you solve this riddle? Vector displacement maps!
But they are also a headache to set up and you have to go into the registry for MR and set it up for it to work properly.
But here’s the result! ONE map but three times the information to make sure your vertices go exactly where you want them!
Milestone 2 has passed!
I spent the week doing a lot of revisions. Turns out getting something past the “looks good enough for games” to “this looks like a real person” is more complicated then I thought. Spent a lot of time pushing and pulling stuff back and forth because I was uncertain on how to proceed. This is the first time that I’m going down to the skinpoores detail-level-wise and skin is REALLY complex!
Anyways! My aim was to present a non-post edited image with an SSS-shader and here it is!
Now I need to sort out proper roughness maps and finish texturing. I know a lot of the skin will pop when it gets some “peach fuzz”, meaning the tiny flimsy hairs all over your body.
DOWN THE TRENCHES BOYS!
It’s all been about planning and sculpting. But mostly sculpting.
So this is the face that I’ll be staring at for the next couple of weeks. On monday I’ll start retopoing, baking and texturing. The goal for next friday is to be able to show a non-postedited still image of a textures face with an SSS-shader.
Here’s the concept sketch from the proposal and the black and white images are from zbrush.
This is the blog for Arvid Burstroms specialization project entitled Avoiding 3D-Botox