Veo 3.1, from Google, is purchasable on 9 of the 11 routes we track. One 8s 1080p t2v clip, with audio costs $2.56 at APIMart, the cheapest published rate we can cite. The first-party route, Gemini API, is $3.20 for the same clip. This model is billed at 8s / 1080p / t2v / audio, which nothing else on the index sells, so its price is not ranked against the other models.
Built around stitching rather than a single shot: chains of scenes extended from one another, with audio generated natively and never optional on the first-party route. That last part is why its price sits outside the comparison the other models share.
The figure above buys one 8s 1080p t2v clip, with audio. The board below prices 3 units in all, each column named. Figures come from each seller's own published rates and carry the date we last read them.
Which route should you take
Each row is the table above read for one question, at the unit named beside the price. Rows marked our call are editorial judgement instead, and appear only where we have something specific to say.
Each column is one unit, and every price in it buys the same thing. The figures quoted elsewhere on this page are one 8s 1080p t2v clip, with audio. That is a baseline for comparing providers, not a quote. What you are actually billed depends on the settings you send, and on some routes on how much you top up at once. Every figure is the provider's own published price; anything we derived rather than read off a price list is marked est.Veo 3.1 generates audio, and the clip compared here includes it.
8 providers tracked · sorted by status, then price
These prices are real, but what they buy is not what their column says, so they are shown and never used to decide which route is cheapest.
CometAPI · 4s/720p
CometAPI does not state whether audio is included at this rate
CometAPI · 8s/1080p
CometAPI does not state whether audio is included at this rate
CometAPI · 8s/4K
CometAPI does not state whether audio is included at this rate
How these prices were read
How each figure above was read off the provider's own price list, and what that price list does not say.
Gemini API
Per-second rate card, no login required. No audio toggle exists for Veo 3.1 on this surface, so the published rate is the only rate.
Vertex AI
Veo 3.1 is documented on the Vertex AI Model Garden, but the generic Vertex generative AI pricing page lists only Gemini text models, not Veo. No public per-second rate card for Veo on Vertex AI has been located.
fal.ai
fal.ai splits audio out as a separate charge; the columns on this board use the with-audio rate, to match the clip this page is priced at. The silent route it also publishes is tracked separately.
Replicate
Replicate publishes per-second rates on the model page, with audio the only modifier it prints and no separate rate by resolution. An earlier note here said no rate was rendered on the page, which described how it had been fetched rather than what the page publishes.
APIMart
APIMart sells Veo 3.1 two ways. The official-channel variants bill per second and state the resolution and whether audio is on, which is what the figures here are multiplied out from. The ext variants bill a flat rate per call and their rows say nothing about clip length, so nothing on the board can be tied to them. It also sells cheaper fast and lite variants of the model, which are not the same product as Veo 3.1 and are not priced here. Every rate is shown beside a standing official price it discounts from, and APIMart bills in credits with the dollar figures printed as approximate.
CometAPI
CometAPI's page does not state whether audio is included in this rate; treated as the closest available match to the clip priced here.
EvoLink
No public Veo 3.1 pricing page has been found on evolink.ai, and the model listing page returns a 404.
kie.ai
kie.ai blocks automated page fetches (403 on kie.ai/veo-3-1); presence and rate are not independently verified.
Veo 3.1 pricing trapsGoogle's own Gemini API and Vertex AI give Veo 3.1 no audio toggle at all: audio is generated natively and is always on, so the published default rate is the only rate on the first-party route. Aggregators are not so strict. fal.ai meters the silent route separately, at $1.60 for the clip this page prices at $3.20 with audio. The board uses the with-audio column throughout, because that is the only thing the first-party route sells.
What it can do
Resolutions
720p, 1080p, 4K
Clip length
4 to 8 seconds
Reference images
up to 3
Reference videos
not verified
Reference audio
not verified
Generates audio
yes
Spoken languages
not verified
Extends a video
yes
Edits a video
not verified
Where a row says not verified, we have not checked that figure yet; it does not mean the model lacks the feature. Every row is read off published documentation for the model, never from our own testing.
Use it without an API key
3 subscription apps sell Veo 3.1
These are not API routes. Each one sells a month of the whole app, so the figure is a subscription and not the price of one Veo 3.1 generation. We record one thing per app: the cheapest plan that unlocks this model, at its monthly-billing rate, and the date we checked. What a generation costs inside a plan is not published anywhere, and dividing a plan by its allowance would invent it.
Veo 3.1 is not sold to consumers per clip. It is bundled into paid Google AI plans, and the tier that unlocks Veo 3.1 itself is Google AI Ultra. Google renders those plans’ monthly figures with client-side script, so no subscription price is recorded here rather than one taken second hand. Every path to the model needs either a paid Google plan or a metered API key: there is no free tier.
Syntx sells a single tier rather than a ladder, so this is both the entry price and the only price. The plan list is visible only when signed in, so the figure is what the plans page showed on the date beside it. The figure is the monthly-billing price, not the lower per-month rate an annual plan advertises, so it is comparable with the other apps here. No per-generation cost is published and we do not derive one: a plan divided by an allowance is an invented price.
Magnific
Magnific, formerly Freepik, sells one paid tier and puts every model it carries inside it. Seedance 2.5 is the exception on this platform: it is not in the model list at all. The figure is the monthly-billing price, not the lower per-month rate an annual plan advertises, so it is comparable with the other apps here. No per-generation cost is published and we do not derive one: a plan divided by an allowance is an invented price.
Higgsfield
Only the entry Starter tier is limited, and it stops at Seedance 2.0; every other model needs Plus. The figure is the monthly-billing price, not the lower per-month rate an annual plan advertises, so it is comparable with the other apps here. No per-generation cost is published and we do not derive one: a plan divided by an allowance is an invented price.
If this is not the right model
When to reach for another model
Veo bills eight seconds at 1080p with audio, which is not a unit anything else here sells, so its figure is not ranked against the others on the index. For a clip that can be compared directly, the Seedance versions are all priced at ten silent seconds of 720p. For speech in named languages, Kling 3.0 generates it natively.
Consumer subscriptions added: Syntx at $17.96 a month, Magnific at $20 and Higgsfield on its $59 Plus tier. Google's own plans still carry no figure, because the price is rendered client-side.
Board gained a 4K column. Google restricts 4K output to the 8-second option, so the 4K clip priced here is the same length as the reference one.
Priced from Google's own Gemini API rate card ($0.40/s standard tier, 720p/1080p, audio included by default) and fal.ai's published per-second table. CometAPI and APIMart publish lower reseller rates. Vertex AI's own per-second pricing could not be located on any public Google Cloud page, and kie.ai blocked automated verification (HTTP 403), so both carry no price yet.
Veo 3.1 released by Google DeepMind: native audio, up to three reference images ('Ingredients to Video'), first/last-frame interpolation, and an Extend feature that chains generations into a single video of roughly 148 seconds. A single generation is still capped at 4, 6 or 8 seconds, and 1080p / 4K output is only available at 8 seconds.