FAQ

Is there going to be a Poser for sampling novel hand and face poses?

The current experiments with VPoser only cover SMPL body pose, that is 63 elements of the full pose vector. For hands the only available option is to use the MANO zero centered PCA space for sampling new poses. For the face pose currently there are no available priors.

---

Is there a way to sample random body shape parameters (e.g. long/short, thin/fat) as well?

The body shape of SMPL(X) is also a centered PCA space that could be potentially sampled from. By keeping the samples within 3-sigma deviation one would most probably get random natural human shapes. Alternatively, one can use Virtual Caliper to create only long/short, thin/fat or even metrically accurate avatars from 3D measurements.