Veo 3.1

Veo 3.1 is Google DeepMind's current video generation model and, since 30 June 2026, the only Veo line the Gemini API serves. It was introduced on 15 October 2025 as a paid preview and kept its predecessor's price. What it added over Veo 3 was control rather than raw fidelity: richer native audio covering natural dialogue as well as synchronised sound effects, a better grasp of cinematic styles, stronger prompt adherence when starting from an image, generation guided by up to three reference images ("ingredients to video"), transitions built from a specified first and last frame, and scene extension — chaining newly generated clips onto an existing video to reach a minute or more of footage. The model has grown through the API changelog rather than through relaunches. On 13 January 2026 Google added 4K output and extended portrait framing to every resolution, so the model now covers landscape 16:9 and portrait 9:16 at 720p, 1080p and 4K, with clips generated eight seconds at a time and stitched further by extension. On 31 March 2026 a cheaper sibling appeared, veo-3.1-lite-generate-preview, aimed at rapid iteration and high-volume applications. Google's product page also lists camera motion control, object insertion and removal, and character consistency carried across shots by reference image. Access runs through the Gemini app, Google AI Studio, the Gemini API, Vertex AI and the Flow filmmaking tool; every generated video carries a SynthID watermark for AI-content detection. One caveat about naming: the API still exposes this as a preview model ID, veo-3.1-generate-preview, alongside the Fast and Lite variants, so it is a production-grade product served under a preview label — Google warns that preview models can change and carry tighter rate limits. The generation before it is gone: Google announced on 15 June 2026 that veo-2.0-generate-001, veo-3.0-generate-001 and veo-3.0-fast-generate-001 would be shut down on 30 June 2026, pointing users to the Veo 3.1 IDs. The Veo 3 profile stays in this catalogue as the earlier version, with that shutdown recorded in it. Google does not publish a parameter count, architecture description or training-data disclosure for any Veo release.

▸Videos

No videos yet.

▸Reviews