OmniHuman

Experimental early-stage avatar-generation model (“Omni Avatar”) that animates human figures from audio input. It focuses on generating vivid talking-head or full-body videos with realistic lip-syncing and facial expressions driven by the supplied speech.