Gemini Omni 1.2 Flash Reference to Video builds a new scene from references: 1 to 10 images of characters, products, props or places, and optionally a reference video of 3 s or less whose motion and camera path it follows. The prompt is optional and describes the action; subjects keep their design and relative size. Native audio including dialogue. Duration 3 to 15 s, 16:9 or 9:16, 360p, 720p, 1080p or 4K. Use the base model instead when you need a first or last frame.