Google announced Veo 3.1 on Wednesday, a video generation model it says builds on May’s Veo 3 with richer audio, better prompt adherence and what it calls enhanced realism that captures true-to-life textures. Veo 3.1 and Veo 3.1 Fast are available in the Gemini app, the Gemini API and Vertex AI, covering text-to-video and image-to-video in horizontal and vertical formats.

The more consequential change is in Flow, the company’s AI filmmaking tool. Audio generation now extends to Ingredients to Video, which uses multiple reference images to lock characters, objects and style; Frames to Video, which bridges a supplied first and last image; and Extend, which continues a clip from its final second and can push a shot past a minute. Flow also gains an Insert tool for adding objects, characters or details into a scene, with Google handling shadows and scene lighting; an object-removal feature that reconstructs background is promised soon.

Google says Flow users have generated more than 275 million videos since the tool launched in May — a figure that does as much work as the model card. Ingredients to Video, first-and-last-frame and scene extension are coming to the Gemini API; scene extension is slated for Vertex AI. Add object and remove object are not available to developers yet.