Google launches Veo 3.1 and new editing features in Flow

Google has unveiled Veo 3.1, a refreshed version of its video generation model, and has brought richer audio and finer-grained editing to Flow, its AI filmmaking interface.

The model is in paid preview via the Gemini API and is also accessible through Vertex AI, the Gemini app and Flow. Pricing has remained the same as Veo 3, according to Google.

What is new in Veo 3.1?

Veo 3.1 has delivered upgrades across audio and control: it has generated more realistic native audio (dialogue and effects), improved prompt adherence, and better outputs when turning images into video, with stronger character consistency across shots.

Three creation tools have been expanded with audio: Ingredients to Video (up to three reference images to steer style or characters), First and Last Frame (Flow’s “Frames to Video”) to bridge two images with a generated shot, and Scene Extension to chain clips for “a minute or more” while maintaining continuity.

Flow has gained precision edits

Flow has added richer, model‑generated audio across its core creation modes and has introduced more precise scene editing.

Creators have been able to insert new objects that respect shadows and lighting; Google has said a remove‑object tool is “coming soon”, reconstructing backgrounds to eliminate artefacts. The company has also said more than 275 million videos have been generated in Flow since its launch five months ago.

Availability and cost

Veo 3.1 and Veo 3.1 Fast have been released in paid preview through the Gemini API (Google AI Studio) and are available to enterprise customers in Vertex AI, with access in the Gemini app and Flow. Google has stated that Veo 3.1 is priced the same as Veo 3.

Google highlighted Promise Studios (using Veo 3.1 within its MUSE platform) and Latitude (for instant story visualisation) as examples of how developers have already incorporated the update.

What is Flow?

Flow is a creative interface powered by Veo focused on directing and editing AI‑generated scenes. It targets filmmakers and creators who need reference‑guided generation, shot‑to‑shot continuity and integrated audio, and it has now adopted more precise object‑level edits alongside longer, chained shots.

The new release has kept Google competitive in fast‑moving AI video. Demis Hassabis, co‑founder and CEO of Google DeepMind, positioned Veo within Google’s broader multimodal strategy; official briefings also framed Veo 3.1 as a step towards more controllable, physically coherent video generation.