Overview

Veo 3.1 shipped in November 2025 and is still the current version as of August 2026. Text-to-video and image-to-video, and unlike most of the field it generates audio natively in the same pass instead of bolting a separate model on afterwards. Output goes to 720p, 1080p or 4K, landscape or vertical. The control surface is decent: reference images through Ingredients to Video, first and last frame transitions through Frames to Video, object insertion and removal, plus camera moves. Access runs through the Gemini API, Vertex AI, AI Studio, Flow, Google Vids and the Gemini app.

Pricing is per successful output second and it is one of the clearer schedules in the category. Standard sits at $0.40 a second for 720p or 1080p and $0.60 for 4K. Fast drops to $0.10 at 720p and $0.12 at 1080p. Lite goes to $0.05 and $0.08. Audio is in the price and failed generations are not billed. An 8-second standard clip is $3.20, which reads fine until you account for the prompts that miss. In practice a finished 8-second shot costs several times that once retries are counted, and Lite at $0.05 a second is the sensible place to burn attempts before committing to a Standard render.

The right pick for a team already on Google Cloud that needs video generation inside a product, with IAM, regional endpoints and an audit trail attached. Marketing groups producing short vertical spots get the most obvious value. The wrong pick if you need shots longer than 8 seconds at full resolution, or if the real task is modifying footage you already have, which Runway's Aleph line and a proper editor handle better. Worth watching whether Google raises the native duration, because that single limit shapes almost every pipeline being built around the model right now.

Key Features

✓ Native audio generated in the same pass as the picture: dialogue, ambient beds and sound effects, synced

✓ 720p, 1080p and 4K output in both 16:9 and 9:16

✓ Scene extension in roughly 7-second increments, chainable past two minutes at the cost of dropping to 720p

✓ Ingredients to Video and Frames to Video for reference-image control and first-to-last-frame transitions

✓ Object insertion and removal plus camera moves, all exposed through the Gemini API, Vertex AI, AI Studio, Flow and Google Vids

My Take

Sora's exit changed the calculus here. OpenAI is pulling its video API on 24 September 2026 with no announced successor, which leaves Veo 3.1 as the default general-purpose video model for anyone who wants a hyperscaler contract behind it. Audio is the reason to pick it: dialogue and effects generated in the same pass, included in the $0.40 per second standard rate rather than billed as a second service. Length is where it frustrates, because the native cap is 8 seconds a prompt and everything past that is a chaining pipeline that drops you back to 720p.