What is Veo 3.1?
Veo is Google DeepMind's video generation model. Google announced Veo 3.1 on October 15, 2025 with richer audio, more narrative control and more realism. The same post describes making a video from several reference images, bridging a start and an end frame, and extending a clip from its last second.
A January 2026 update added vertical 9:16 video from reference images, and 1080p and 4K output through upscaling.
Fast, Quality and Lite
Google's Vertex AI documentation lists three Veo 3.1 models. All three make 4-, 6- or 8-second clips, generate sound, and take a first and last frame. The page lists the two main models as generally available from November 17, 2025 and Lite as a preview from April 2, 2026.
9to5Google reported that Lite costs less than half as much as Fast and runs at the same speed.
On MakeMidas, Veo 3.1 Quality makes 8-second clips at up to 4K, always with sound. Fast and Lite make 4-, 6- or 8-second clips at 720p or 1080p, with sound you can switch off. All three start from text, a first frame, or a first and last frame.













