What to look for
This ten-second vertical 720p OmniHuman 2.0 clip was generated from a single photo and a speech recording, with no editing. The lips match the words, her other hand gestures, her head tilts in the pauses and the bottle stays in her hand to the end. The portrait was generated by AI and the voice is synthetic: no real person appears in the clip. The model page shows two more examples: a storyteller with gestures and a cartoon character with emotions.