Gemini Omni

New multimodal video generation model, that can also create custom avatars from your own voice and likeness.

Takes image, audio, video, or text as input and produces editable video output
Lets you create an AI avatar from your own voice and likeness, so videos look and sound like you
Editing happens through normal conversation: describe a change and Omni applies it
Every Omni video carries a hidden SynthID watermark, and a more capable Pro version is coming