i mean, we don’t know what the training was like for this version. if it was trained on a limited subset of people that could easily explain why. In future games that want to support this, they could train it specifically on the actor/models they used.