Out of curiosity, what underlying models are you using for the lip-syncing feature? Does it handle multiple speakers well?

Portrait to video