Veo 3 is a generative video model from Google (DeepMind) that produces synchronized audiovisual outputs from text or image prompts. It creates high-fidelity video (commonly demonstrated at 1080p) combined with native audio including dialogue, sound effects, and ambient noise, enabling single-request generation of complete clips. Veo 3 is exposed via production APIs (Vertex AI / related endpoints) with multiple variants (veo3, veo3-pro, veo3-fast, veo3-pro-frames) to balance quality, speed, and frame controls. The model includes safety filtering and imperceptible watermarking and is designed to integrate with creative interfaces (e.g., Flow) that provide fine-grained camera, motion, and perspective controls for filmmaking-style workflows.