Veo 3.1 API and Pricing, Explained Through a Real Pipeline
Veo 3.1 prices by the second, and the number changes with resolution and whether you keep the native audio on. Here is the exact per-second pricing, worked examples at real clip lengths, and how to reach the model through a published pipeline API instead of building the integration yourself.
Veo 3.1 is Google's flagship video model, and it is the one most teams reach for when a shot has to look expensive: film-like lighting, camera moves you can direct in plain language, and dialogue, ambience, and sound effects generated in sync with the picture rather than layered on after. What trips people up is the pricing. Veo 3.1 does not charge a flat rate per clip. It charges per second, and the number depends on resolution and whether audio is on, so two people describing "the cost of a Veo clip" can be talking about numbers three or four times apart. Here is the actual pricing, worked through real clip lengths, and how to call the model as a published API without building the Google integration yourself.
What Veo 3.1 costs per second
Veo 3.1 supports clips of 4, 6, or 8 seconds, at 16:9 or 9:16, up to 4K. The per-second rate on Treza's pipeline (the same metered cost the model charges upstream, no markup hidden in a subscription) breaks down like this:
| Setting | Price per second |
|---|---|
| With audio, standard resolution (720p/1080p) | $0.40 |
| With audio, 4K | $0.60 |
| Without audio, standard resolution (720p/1080p) | $0.20 |
| Without audio, 4K | $0.40 |
Two things jump out. First, audio roughly doubles the price at every resolution: generating synced dialogue and ambience is real additional compute, not a toggle. Second, 4K costs 50 percent more than standard resolution at the same audio setting. If you are rendering for a vertical feed where nobody is watching at 4K, standard resolution is not a compromise, it is the correct setting.
What that means at real clip lengths
Pricing per second is only useful once you multiply it by a duration you'd actually ship. Veo 3.1 only accepts 4, 6, or 8 second clips, so here is the full grid:
| Duration | With audio, standard | With audio, 4K | Without audio, standard |
|---|---|---|---|
| 4 seconds | $1.60 | $2.40 | $0.80 |
| 6 seconds | $2.40 | $3.60 | $1.20 |
| 8 seconds | $3.20 | $4.80 | $1.60 |
An 8-second hero shot with native dialogue at standard resolution runs $3.20. Drop audio and it is $1.60. Push to 4K with audio and the same 8 seconds is $4.80. None of that is charged as a subscription line item: on Treza it is metered from a prepaid credit balance, where 1 credit equals $0.01, packs start at $5, credits never expire, and a failed generation charges nothing. There is no minimum spend and no monthly forfeiture if a week goes by without a render.
The cheap way to iterate: Veo 3.1 Fast
Nobody lands on the right prompt on the first try, and paying full Veo 3.1 rates to test five framings of the same shot adds up fast. Veo 3.1 Fast is the same model family, same native audio, at a fraction of the cost: $0.12 per second with audio at standard resolution, versus $0.40 on full Veo 3.1. That is roughly a third of the price for the exploration loop. An 8-second draft with audio on Fast costs $0.96, against $3.20 on the full model.
The pattern that keeps a production budget sane is drafting on Fast and re-rendering only the winning prompt on full Veo 3.1. Fan a brief into a few prompt variants, render all of them on Fast, pick the one that works, and spend the full-quality render budget once instead of on every attempt.
Calling Veo 3.1 as an API without building the Google integration
Getting direct API access to Veo 3.1 means a Gemini API key or a Vertex AI project, handling an asynchronous submit-and-poll job pattern yourself, and building fallback logic for when a request gets rate limited or rejected by content moderation. That is real integration work before you have written any of your actual product logic.
Through a Treza pipeline, Veo 3.1 is one node on a canvas rather than a standalone API you integrate against directly. You place the node, set duration, resolution, and audio on or off, wire it next to a script node, a voiceover node, or a captioning step, and publish the whole chain as a single versioned endpoint. Your app calls the typed /invoke route with JSON in and JSON out, or points an existing OpenAI SDK at the OpenAI-compatible route if that is what your stack already speaks. Swap the model on the node later, drop in Sora 2 Pro or Seedance 2.5 for a different shot, and the endpoint your product calls does not change.
The same pipeline is also reachable from an agent. Treza's hosted MCP server exposes pipelines as tools Claude and other MCP clients can call directly, so "render this brief on Veo 3.1 and give me the link when it's done" is a prompt an agent can actually carry out, not just describe.
First and last frame control
One Veo 3.1 feature worth knowing about before you price a project: it accepts a first-frame and a last-frame image and generates the motion between them. That is useful for anything brand-locked, a product reveal that has to open on an exact hero shot and close on an exact logo card, without leaving the framing to the prompt alone. It works at every resolution and audio setting in the pricing table above, so bookending a shot does not change the cost.
Putting it in a pipeline instead of a single call
A single Veo 3.1 render is rarely the finished deliverable. A real video pipeline usually chains a script step ahead of the video node and narration, transcription, or caption steps after it, so the video model handles the shot while the rest of the graph handles everything around it. Built once, that chain reruns on a schedule or from an API call without anyone touching the canvas again, and the render cost of the Veo 3.1 node is just one line item in whatever the full run charges.
If you are starting from a blank canvas, the AI video generator page has a template that puts Veo 3.1 (or any model in the catalog) behind a single generate action, and you can wire additional steps around it once the first shot looks right.
Frequently Asked Questions
How much does Veo 3.1 cost per second?
Standard resolution with audio is $0.40 per second, standard resolution without audio is $0.20 per second, and 4K adds a further 50 percent on top of whichever audio setting you use, up to $0.60 per second with audio at 4K.
How much does a typical Veo 3.1 clip cost?
An 8-second clip, the longest Veo 3.1 supports, costs $3.20 with audio at standard resolution, $1.60 with audio off, or $4.80 with audio at 4K. Shorter 4 and 6 second clips scale down proportionally.
Is there a subscription required to use Veo 3.1 through an API?
Not through Treza. Veo 3.1 renders are charged from a prepaid credit balance, 1 credit equals $0.01, packs start at $5, and credits never expire. There is no monthly plan and no forfeiture if you don't generate for a while.
What's the cheapest way to iterate on Veo 3.1 prompts?
Draft on Veo 3.1 Fast, which runs $0.12 per second with audio at standard resolution instead of $0.40, and re-render only the winning prompt on full Veo 3.1 once you know the shot works.
Can I call Veo 3.1 without setting up a Google Cloud or Gemini API account myself?
Yes. On a Treza pipeline, Veo 3.1 is a node you configure once and publish as a versioned HTTP endpoint, with a typed /invoke route and an OpenAI-compatible route, so your product calls one API regardless of which model is behind the node. A hosted MCP server also exposes the same pipelines as tools for Claude and other agents.
Does Veo 3.1 support 4K output?
Yes. Veo 3.1 renders at 720p, 1080p, or 4K, with 4K priced at a 50 percent premium per second over standard resolution at the same audio setting.
