Despite thinking a few weeks ago that I had worked out the facial rigging system, the finer points of it, and my sanity temporarily, all said goodbye. To cut a very long story short, the rock-solid process that worked on the first model, didn’t on the second. After many a long night trying to understand why, and messaging a few key people, I rolled back to an older version of the Advanced Skeleton rigging system that I use after not being able to pinch in the mouth of the second model. It all works fine now, thankfully. As it also happens, we have been granted an extension to our deadline which is a huge relief, it’s a shame I was not able to meet the original deadline and post the project online. I felt that digging deep and understanding facial rigging would be better in the long run, rather than use off the shelf models that didn’t necessarily tell my story.
So, firstly I had to neutralize the face to get allow for a broader range of expressions. This involved using the grab tool on the Maya sculpting bar, this was the first part of a super useful tutorial that Amedeo produced for me, he has been an invaluable source of technical help over the past two weeks, patiently and with immense technical depth, resolving my issues one by one. Here’s the nose frozen so I don’t distort it as the face is pulled down. Also, whilst going through this process, you need to smooth out the edges and topology for good deformations.
You can see by the results here that the smile and eyebrows have been lowered, this will allow for a wider range of emotional expression.
After resolving this I wanted to see what he looked like in Unreal so did a few FBX and Alembic import tests, largely to make me feel better about building a set and getting things staged and lit nicely. It was very satisfying to see some kind of vision take shape.
After this I ran into all sorts of difficulties with the lips not pulling in as I described earlier. I sent an image of the problem to Øyvind Nøstdal, the super clever guy behind the rigging setup. I did remember from one of his facial tutorial videos, that you can see which version was used to create the rig within a file so I dug into the Conan rig and saw that it was built in previous version, that rig had the right deformations for a wide mouth so I managed to employ it just fine. You can see from the comparisons in this image below, how the mouth should have been pinching together.
As it wasn’t working, I went through so many different possible solutions; cutting the model in half to re symmetrize, rebuilding it in different slight variations. Even after I employed the old face setup, I realised after building it that it was super slow in Maya because I’d used one version for the rig and one for the face. Eventually after getting some more expert mentoring with weight painting, learning to always add weights, never subtract, some NGSkin smoothing and mirroring, he’s looking pretty good, really flexible and expressive.
And now for some animation…