Veo 3.1

What it is

Veo 3.1 is an update to Veo 3, released on 15 October 2025 ^[2]. Google DeepMind describes Veo as a model for cinematic video with audio ^[3]: the output is a moving image sequence with a soundtrack, not silent footage. Veo 3.1 keeps that framing and adds controls aimed at longer and more carefully staged work.

Where it is used

Veo appears in two Google products with different jobs. In the Gemini app it is used for shorter projects; in Google Flow it is used for longer projects that need continuity ^[2]. Continuity is the practical difference. A short clip can be generated on its own, while a longer sequence has to keep characters, objects and style recognisable from shot to shot, and Flow is built around that requirement.

The API versions

On the same day, 15 October 2025, the model appeared in the Gemini API under two names, veo-3.1-generate-preview and veo-3.1-fast-generate-preview ^[1]. Two names for one model usually signal a trade-off between quality and speed: one leans towards the finished result, the other towards faster generation. A lower-cost veo-3.1-lite-generate-preview followed on 31 March 2026, adding a cheaper tier below the first two ^[1]. The word "preview" in all three names marks them as models that may change or be withdrawn, unlike a stable release.

Capabilities announced with the release

Google said Veo 3.1 brought audio to Flow's existing capabilities for the first time, and that Flow supports audio across all of its features ^[4]. Two capabilities were singled out. Ingredients to Video lets a user provide multiple reference images to control characters, objects and style ^[4]; reference images act as visual instructions, so the same character or object can stay recognisable across generations. Extend can create videos lasting a minute or more, with each new clip generated from the final second of the previous clip ^[4]; chaining clips this way stretches a limited generation window into a longer sequence without a hard cut.

Some of these capabilities were also on their way into the Gemini API: Ingredients to Video, First and last frame, and Scene extension ^[4]. Add object and Remove object were not yet available there ^[4].

Where it sits in the line

Veo 3.1 is a point update rather than a new generation: the number carried over from Veo 3, and the release concentrated on audio, reference images and clip extension rather than on a new model family ^[2]. Its practical successor in Google's own guidance turned out to be a different model altogether, gemini-omni-1.2-flash ^[1].

Retirement of the previews

All Veo 3.1 API previews were shut down on 22 October 2026, and Google recommended gemini-omni-1.2-flash instead ^[1]. Slightly more than a year separated the first preview from its retirement, a short lifespan that follows from the preview label: these names were never meant to be permanent endpoints for developers to build on.