What it is
Veo 3.1 is no longer available on the Gemini API. Google shut down all three preview model IDs (veo-3.1-generate-preview, veo-3.1-fast-generate-preview and veo-3.1-lite-generate-preview) on October 22, 2026, and names Gemini Omni Flash as the replacement there. The model lives on in Google Cloud's Gemini Enterprise Agent Platform, the renamed Vertex AI, where veo-3.1-generate-001 and veo-3.1-fast-generate-001 have been generally available since November 17, 2025, with a listed retirement date of November 17, 2026 or later. Lite is still a preview there. In the Gemini app, Google says Omni 1.1 Flash replaces Veo 3.1; Flow's help centre still refers to both Veo and Omni.
Veo 3.1 and Fast were announced on October 15, 2025; Lite followed on March 31, 2026. All three make 4, 6 or 8 second clips at 24 frames per second in 16:9 or 9:16, with or without generated sound on Google Cloud. The standard and Fast models reach 4K on Google Cloud's price list; Lite stops at 1080p.
The model takes up to three reference images (not on Lite), accepts a first and last frame, and can extend a Veo clip by 7 seconds at a time, up to 20 times, to about 148 seconds at 720p. Outputs carry SynthID watermarks, and Google Cloud lists C2PA Content Credentials support.
Our review
Veo 3.1 is a model you keep running, not one you adopt. The October 22 shutdown removed it from the Gemini API, which is where most developers met it, and left it on Google Cloud's enterprise platform in a single US region with a listed retirement date of November 17, 2026 or later. Anyone starting a video project on Google's stack today is being pointed at Gemini Omni Flash.
The leaderboard explains why. On Artificial Analysis's text-to-video board the standard model sits twenty-first at 963 Elo, with Fast at 960 and Lite at 950, all below the Kling 3.0 anchor at 1000. Gemini Omni Flash 1.1 scores 1130 and Wan 3.0 leads at 1175. Fast and standard are statistically level, so paying $0.40 a second for the flagship over Fast's $0.12 at 1080p buys 4K and little else.
What Veo 3.1 still offers that Omni does not document is length. Omni Flash 1.1 can extend clips, but Google publishes no maximum; Veo's 7-second steps to roughly 148 seconds, explicit last-frame control and three reference images are documented and stable, and Google's own video docs once named these as the reasons to pick it. Google Cloud also adds things the Gemini API never had: a video-only price at half the audio rate, up to four outputs per prompt, customer-managed encryption keys and VPC Service Controls.
The catch for a migrating team is the contract, not the code. Google Cloud lists the GA models with Provisioned Throughput and fixed quota but no batch inference, and the quota is 50 requests a minute per model in us-central1. A team that moves there to keep Veo is buying time on a model that ranks below twenty others on the same board.
Where it earns its place
A backend engineer whose app called veo-3.1-fast-generate-preview on the Gemini API
Keep long, extended clips working after the October 22 shutdown
Move to veo-3.1-fast-generate-001 on Google Cloud in us-central1, with a service account instead of an API key. Gemini Omni Flash is the cheaper switch and does extend clips, but Google publishes no maximum length for it, so long chained extensions need testing before the move.
A video producer at an agency on Google Workspace
Storyboard a 2-minute sequence with a fixed start and end shot
Generate an 8-second 720p base clip, set the last frame, then extend it in 7-second steps up to 148 seconds. At Fast's $0.10 a second with audio, 120 generated seconds cost $12; Google does not say how extensions are billed, and Lite takes no reference images.
A procurement lead at a regulated media company
Buy AI video under a Google Cloud contract rather than a consumer API
Veo 3.1 GA comes under Google Cloud's terms: no training on inputs without permission, CMEK and VPC Service Controls. The trade-off is a model Google has already replaced on its developer API, with retirement listed from November 2026.
Before you commit
- Setup
- Needs a Google Cloud project with the Agent Platform API enabled in us-central1; Gemini API keys no longer reach any Veo 3.1 model. Generation is asynchronous and the old Gemini API docs put latency at 11 seconds to 6 minutes.
- Integrations
- Google Gen AI SDK and REST on Google Cloud, Agent Studio, and Google Flow for Google AI subscribers. The Gemini app now uses Omni 1.1 Flash instead.
- Data and privacy
- Google Cloud says it does not train on customer data without permission, for GA and pre-GA models alike; zero data retention needs extra configuration. Outputs carry SynthID, and Content Credentials (C2PA) are supported.
- Cost at scale
- 1,000 eight-second 1080p clips with audio cost about $960 on Fast, $3,200 on the standard model, or $1,600 on standard without audio; Lite at 720p without audio is $240. Provisioned Throughput is offered; batch inference is not.
- Limits
- Clips of 4, 6 or 8 seconds, up to four per prompt; English prompts only; 50 requests a minute per model in us-central1. Retirement on Google Cloud is listed as November 17, 2026 or later.
What's good
- Documented extension to about 148 seconds at 720p
- First-and-last-frame control and up to three reference images
- GA on Google Cloud with CMEK, VPC-SC and no training on customer data
- Video-only option at half the with-audio price
What's not
- Shut down on the Gemini API on October 22, 2026
- Twenty-first on Artificial Analysis's text-to-video board, below the 1000 anchor
- Google Cloud availability limited to us-central1, retirement listed from November 17, 2026
- Standard tier costs $24 a minute at 1080p with audio
Best for
Teams with existing Veo pipelines that depend on long extensions or end-frame control and can run on Google Cloud.
Alternatives
Veo 3.1: common questions
- Is Veo 3.1 worth it?
- Veo 3.1 is now an enterprise-cloud legacy model: still GA on Google Cloud, gone from the Gemini API, and 170 Elo behind its own replacement. We rate it 3.3 out of 5.
- How much does Veo 3.1 cost?
- Google Cloud, per second: Veo 3.1 $0.40 with audio or $0.20 without (720p/1080p), $0.60/$0.40 at 4K; Fast $0.10 to $0.12 with audio; Lite $0.05 to $0.08 with audio, from $0.03 without.
- What are the downsides of Veo 3.1?
- Shut down on the Gemini API on October 22, 2026; twenty-first on Artificial Analysis's text-to-video board, below the 1000 anchor; google Cloud availability limited to us-central1, retirement listed from November 17, 2026; standard tier costs $24 a minute at 1080p with audio.
- Who is Veo 3.1 best for?
- Teams with existing Veo pipelines that depend on long extensions or end-frame control and can run on Google Cloud.
- How much does Veo 3.1 cost for a team or in production?
- 1,000 eight-second 1080p clips with audio cost about $960 on Fast, $3,200 on the standard model, or $1,600 on standard without audio; Lite at 720p without audio is $240. Provisioned Throughput is offered; batch inference is not.
- Is Veo 3.1 safe for company data?
- Google Cloud says it does not train on customer data without permission, for GA and pre-GA models alike; zero data retention needs extra configuration. Outputs carry SynthID, and Content Credentials (C2PA) are supported.
- What are the best Veo 3.1 alternatives?
- The closest alternatives we have reviewed are Gemini Omni Flash 1.1, Wan 3.0, MiniMax H3 (Hailuo 3.0) and Seedance 2.5.
Answers checked Oct 22, 2026.