Veo 3.1 pricing (2027): API rates, Flow credits & cost

You want to budget a Veo video, but the route in an older pricing table may no longer accept new jobs. A historical per-second rate is not useful as a purchase quote if the endpoint behind it has retired.

This guide separates the earlier Gemini API Veo 3.1 prices from the access decisions you need to make now. You’ll see how direct API billing differs from Flow credits and where model choice changes the comparison. Choose the platform and live endpoint first, then use the dated calculations to understand the budget rather than assuming the old IDs remain available.

Watch Veo 3.1 examples

Google DeepMind’s Veo 3.1 demonstration introduces audio and creative controls. These are vendor-selected examples, not our side-by-side test.

Source: Google DeepMind: Veo 3.1 - Designed to empower creatives

Historical Gemini API Veo 3.1 pricing at a glance

Google shut down its Gemini API Veo 3.1 Preview endpoints on October 22, 2026. The dollar rates below were checked September 13 and describe the retired Preview models. For a new direct Gemini integration, Google points to gemini-omni-1.2-flash or available Cloud GA models. Flow uses its own credits and rates.

Google’s current video guide confirms that the Standard, Fast and Lite Gemini API preview IDs were shut down October 22, 2026. That does not establish shutdown of every service offering Veo. Read Google’s migration guidance before starting a new integration, and check the replacement endpoint’s pricing separately.

Google Gemini API pricing page, captured September 25, 2026. Billing selection, currency and any promotion are shown in the screenshot; recheck the live page and checkout before buying. View official source

At the September 13 check, Google said the Gemini API charged only when a video was successfully generated. Those historical billing rules do not establish access to a retired endpoint. Budget for all charged attempts on the current route, then divide the relevant spend by accepted clips.

How much does one Veo 3.1 video cost?

For an eight-second Veo 3.1 request with native audio, multiply the selected variant’s per-second rate by eight. Using the Gemini API rates checked September 13, 2026: Standard costs $3.20 at 720p or 1080p and $4.80 at 4K. Fast costs $0.80 at 720p, $0.96 at 1080p or $2.40 at 4K. Lite costs $0.40 at 720p or $0.64 at 1080p and does not support 4K. Eight seconds is the example here, not the only supported duration.

These figures are arithmetic from the listed rate, not an output-quality comparison. A cheaper attempt can cost more per accepted clip if it requires more retries or correction.

Choose duration and input before the cheapest rate

Historical request settings, recorded September 13, 2026: Google’s current Veo API parameter guide listed 4-, 6- and 8-second generation options, with eight seconds required for 1080p, 4K and supported reference-image requests. At the rates shown then, a four-second 720p request would have cost $0.20 with Lite, $0.40 with Fast or $1.60 with Standard. Check the chosen provider’s controls before quoting a new job.

Download approved API videos within two days. Google’s Veo guide says generated videos are stored on its server for two days before removal. Retrieve the completed file into your own storage before that window closes; an API result is not a permanent archive. Extended videos are treated as newly generated videos. This retention rule belongs to the Gemini API guide and should not be substituted for another provider’s storage policy.

Veo 3.1 Standard vs Fast vs Lite

• Standard: the earlier 720p and 1080p rate was $0.40 per second, or $0.60 at 4K.

• Fast: the earlier rates were $0.10, $0.12 and $0.30 per second at 720p, 1080p and 4K.

• Lite: the earlier rates were $0.05 at 720p and $0.08 at 1080p. This variant did not support 4K.

Do not infer that the names alone establish a winner. Run the same brief, input assets, aspect ratio and review rubric across variants. Save every charged attempt and record which clips actually meet the deliverable.

Is Veo still Google's default video model?

Google's current Gemini API video overview recommends Gemini Omni 1.2 Flash for current direct Gemini video generation and editing. Google lists its 720p video output at about $0.07 per second, with input tokens billed separately. If you need a Veo-specific control such as a last frame, verify the available Veo route, model ID and supported settings before building around it. The October 22 Preview retirement applies to the Gemini API IDs, while Google Cloud and other providers have separate model availability.

That distinction matters when comparing prices. A Google AI plan may advertise video access through Gemini or Flow without promising that every request uses Veo 3.1. Check the selected model in the interface or API request.

How Google Flow pricing differs from the API

Google's current plan comparison lists 200 monthly Flow credits for Google AI Plus, 1,000 for Pro, 10,000 for the 5x Ultra tier and 25,000 for the 20x Ultra tier. The page does not expose one universal cash price in every locale, and the number of credits consumed depends on the selected Flow operation. Use the checkout price for your country and the credit quote shown before generation.

A Flow credit is not an API dollar and does not prove a fixed number of clips. Keep consumer-plan invoices, included storage and other benefits separate from an API unit-cost calculation.

A budget for accepted clips

Suppose you need twenty accepted eight-second clips. At one charged attempt per accepted clip, Veo 3.1 Fast at 720p would cost $16 in generation charges. At three charged attempts per accepted clip, the same sixty attempts would cost $48. Standard at 1080p would cost $64 at one attempt each or $192 at three.

The retry counts are scenarios, not measured success rates. Define acceptance before testing: subject identity, product accuracy, motion, camera, audio, text, continuity and legal-use requirements. Report the number of accepted clips and the total relevant spend.

Choose the live route before reusing an old Veo budget

The current Gemini video overview names the retired IDs and links to Gemini Omni Flash and Cloud GA Veo documentation. Before changing an integration, record its host, model ID, input controls and output format. A model-name replacement is not proof that the old request body or price still applies.

Gemini API, Vertex AI and third-party providers

The Gemini Developer API price above is one purchasing surface. Vertex AI and third-party platforms can expose different model IDs, controls, regions, quotas, storage, support and rates. Quote the provider and interface beside the model name. Never copy Google's direct rate into a third-party invoice estimate.

For programmatic use through Magic Hour, see Veo 3.1 API access for supported settings, code examples and current credit estimates. Magic Hour billing is separate from Google’s direct API rates.

Use our AI video pricing index for other direct examples and the AI video-generator comparison when the decision includes complete creator platforms rather than only model endpoints.

How to compare Veo with another model

Historical rates were checked September 13, 2026 against Google's Gemini API pricing, Veo 3.1 API guide and Google AI plan comparison. Current replacement guidance and Gemini Omni 1.2 Flash pricing were checked November 9, 2026. Dollar amounts are USD before taxes, discounts or provider markups.

• Choose five representative briefs, including the hardest subject, motion, camera and audio requirement.

• Use the same source images, aspect ratio and acceptance rubric where the models support them.

• Save model version, provider, settings, prompt, seed where available, output and charged amount for every attempt.

• Count accepted outputs only after full-speed review at the intended delivery size.

• Compare generation spend, review time, corrections and integration work per accepted clip.

You can also review the current Magic Hour model catalog or start a representative prompt in Text to Video. Availability and credit estimates can change by selected model, so verify the live picker before budgeting.