/3/ - 3DCG

>>965869
Machine learning? Why would I need machine learning? My system will not need to guess where vertices should be based on learned behavior. It will know where vertices should be, because the math will be accurate. It's not predictive, it's deterministic. You get out exactly what you put in.
In fact, I've looked at ziva before. Only a small glimpse admittedly. And hearing how it uses AI is one of the things that annoyed me enough to start my own journey to find my own solution. Because I'm convinced that all the tech nerds are overlooking a more foundational solution to deformation, and instead chasing fancy bell&whistle gimmick solutions.

When(or if) my way is developed, then the user will only have to plot a few points, and draw some lines between the points, and voila, they have a semblance of "muscle" working under their mesh. And it will work in tandem with existing bones. They won't have to create a whole muscle structure either. If they feel only 1 particular area is giving them difficulty, then a single point might do the trick.

I'm getting heated just thinking about how underdeveloped this idea is. It pisses me off. It feels like some egghead should have thought to implement it into basic 3D works back in the 90s. Anyway, check out this image. Using the viewer node on one of the distance nodes, it makes this cool looking visual effect on the body. You can see the zones of influence the empties create.
I just noticed one of my empties were misplaced. Oops, a pelvic empty somehow ended up down near the foot. Fixed that. And oh shit, I just realized that I undid all the parenting of the pose bones, so my previous 2 webms aren't even accurate.(I was really tired) But oh well, it was still accurate enough. The fixed version still looks more or less the same.

>>965874
>>965875
Looks really sexy.