|07 July 2012||#16|
Snr Technical Artist
Join Date: Feb 2002
Data is definitely moving into the computational arena, rather than a manual one - for example solving facial blendshapes in a pose space, rather than hand-crafting and knowing how a mix needs to be properly interpolated. Solving issues computationally has two benefits:
firstly input data can be hand-crafted, and secondly solved data can be 'tuned' aesthetically.
It's not really a case of using fusiform approaches over pose space deformations - its more a case you want to 'solve' the variety of unknowns that the character can do - unknown poses, motions etc that need to hold up on the screen or in the game. And the ability to 'craft' exact shapes for hero frames/shots.
Building a full muscle system is a huge endeavor, and what tends to happen is a combination of a broad muscle system with PSDs and shrink-wrap skin sim's for fine creases.
A solved system is nice in that it can be very finite - i.e only solving for certain situations or very broad i.e every possible combination of a rig, slider bank etc.
Faceware by example is doing this - when the talent makes a smile pose in the video, you make a 'key' smile pose using sliders and faceware then solves for the whole video. The benefit is again tune-ability and the fact that after the data's been solved you can still hand key poses and make refinements.
Disclaimer: My opinions are not those of my employer.
|Thread Closed share thread|