AI video generation API cost in 2026: per second, per clip and per minute

The short answer

Video APIs bill per second of output. At 720p on Revana, prices run from $0.05 a second (Seedance 2.0 Mini, MiniMax H3 and Veo 3.1 Lite) to $0.40 a second for Veo 3.1 with sound, so a 10 second clip costs between $0.50 and $4.00. Resolution, sound and length move the price the most; a job that fails costs nothing.

Price per second, per clip and per minute (720p)

Each row is the model's 720p rate (768p for MiniMax H3), with sound where the model makes it. "1 minute" is 60 seconds of footage, which for most models means several clips.

Every model has more settings (480p, 1080p, 4K, a video input) than one row can show; each model page and the pricing page list them all.

What makes a clip cost more

• Resolution. 1080p is often two to three times the 720p rate: Seedance 2.5 goes from $0.21 to $0.52 a second.
• Sound. Some models charge extra for it: Kling 3.0 is $0.09 a second silent and $0.13 with sound. Others, like Kling 4.0 and Seedance 2.5, always include it.
• Length. The price is per second, so a 30 second clip costs three times a 10 second one. Write long clips only once the look is right.
• A video input. Editing or following a clip has its own rate: Seedance 2.5 is $0.15 a second at 720p with a video input.
• Retries. On Revana a failed job is free, so you pay only for clips you get. Not every API works that way, so check before you compare list prices.

Worked example: a 30 second ad

Say you need a 30 second vertical ad and expect to try each shot twice. That is 60 seconds of generated footage at 720p:

• Wan 3.0: $4.20, a low-cost choice for drafts and b-roll.
• Kling 4.0, with sound: $8.10.
• Seedance 2.5, with sound: $12.60.

A common pattern is to draft every shot on a low-cost model or at 480p, then render only the final takes on the model you like best. Seven ways to cut AI video API costs covers this in more detail.

Estimate a job before you run it

Send the same request body to POST /v1/generations/quote with your API key and Revana answers with the price without making or holding anything. Some inputs cost more than a model's listed rate (Seedance 2.5 with a first frame, for example), so quote the exact request and pass the quote as max_cost_usd: a job that would cost more is refused before anything is charged. In the dashboard, the Generate page shows the price before you press the button. The quickstart has the request shapes.

Comparing with other APIs

This page compares models on one API. To compare the same model across providers (fal, Replicate, Kie.ai and others), see the AI video API price comparison and the cheapest AI video generation API guide, which list other APIs' prices with the date each was checked.

Frequently asked questions

How much does AI video generation cost through an API?

On Revana in November 2026, a 720p clip costs from $0.05 a second (Seedance 2.0 Mini, MiniMax H3 and Veo 3.1 Lite) to $0.40 a second (Veo 3.1 with sound). A 10 second clip runs from $0.50 to $4.00.

How much does one minute of AI video cost?

At 720p, 60 seconds of footage costs $4.20 on Wan 3.0, $8.10 on Kling 4.0 and $12.60 on Seedance 2.5. Most models make clips of 15 or 30 seconds at most, so a minute is several clips.

Do failed generations cost anything?

No. On Revana a job is charged only when it succeeds, and the finished job reports the exact amount in cost_usd.

Is 1080p much more expensive than 720p?

It depends on the model. Seedance 2.5 costs $0.52 a second at 1080p against $0.21 at 720p, while Kling 4.0 goes from $0.135 to $0.20. Draft at 720p or 480p and render the final cut at 1080p.