AI video generation API pricing
What 14 text-to-video and image-to-video models cost through one OpenAI-compatible API, sorted by the price of a 5-second clip. No subscription: you pay per second of finished video from a prepaid balance.
Video model rates
Our live pricing feed is unavailable right now, so these are indicative list rates; the pricing page has current rates. Resolution and latency are what we measured on a 16:9 clip.
| Model | Rate | 5 s clip | 10 s clip | Output | Typical latency |
|---|---|---|---|---|---|
| MiniMax H3 (768p) MiniMax | $0.030 / s | $0.15 | $0.30 | 1344 × 768 (16:9)audio | ≈3.5 min |
| Kling Omni 4.0 Kuaishou | $0.085 / s | $0.43 | $0.85 | 1280 × 720 (16:9)audio | ≈5 min |
| Seedance 2.0 Mini BytePlus (Volcengine) | $0.085 / s | $0.43 | $0.85 | ≈864 × 496 (16:9)audio | ≈2 min |
| Seedance 2.0 BytePlus (Volcengine) | $0.64 / clip | $0.64 | $0.64 | ≈864 × 496 (16:9)audio · image-to-video | 2–4 min |
| Kling O1 Kuaishou | $0.17 / s | $0.88 | $1.75 | 720p (16:9)video only | ≈1.5 min |
| Wan 3.0 Video Prime Alibaba | $0.20 / s | $0.98 | $1.95 | 1920 × 1080 (16:9)audio | ≈2.5 min |
| Seedance 2.0 Fast BytePlus (Volcengine) | $0.25 / s | $1.25 | $2.50 | ≈864 × 496 (16:9)audio | ≈2.5 min |
| Kling 3.0 Omni Kuaishou | $0.26 / s | $1.27 | $2.55 | 720p (16:9)video only | ≈3 min |
| HappyHorse 1.1 Alibaba | $0.31 / s | $1.55 | $3.10 | 1080p (16:9)video only | ≈2 min |
| Seedance 2.0 4K BytePlus (Volcengine) | $0.38 / s | $1.90 | $3.80 | up to 4K (16:9)audio | ≈3 min |
| HappyHorse 1.0 Alibaba | $0.48 / s | $2.40 | $4.80 | 1080p (16:9)video only | ≈2.5 min |
| Seedance 2.5 BytePlus (Volcengine) | $0.55 / s | $2.73 | $5.45 | ≈864 × 496 (16:9)audio | ≈2.5 min |
| Kling 3.0 Kuaishou | $0.56 / s | $2.82 | $5.65 | 720p (16:9)video only | ≈3.5 min |
| Wan 3.0 Video Alibaba | $0.62 / s | $3.10 | $6.20 | 1080p, 30 fps (16:9)video only | ≈2 min |
How video API billing works
- Most models bill per second of finished video, so the clip length you request is the main cost driver. Models billed per clip charge the same for any length they support.
- You are charged when the task is accepted and refunded automatically if it fails. Queueing and render time are free.
- Every model takes the same request:
POST /v1/videoswith a prompt and a length, then poll the task and download the clip. See the video API docs. - Need a monthly estimate? The cost calculator multiplies these rates by your clip length and volume. For a side-by-side with another provider, see TokPort vs fal.ai.
Video API pricing FAQ
How much does an AI video generation API cost?
On TokPort, a 5-second clip costs from $0.15 (MiniMax H3 (768p)) to $3.10 (Wan 3.0 Video), across 14 video models. Most models bill per second of finished video, so a 10-second clip costs twice a 5-second one.
What is the cheapest AI video generation API?
MiniMax H3 (768p) at $0.030 / s, or $0.15 for a 5-second clip at 1344 × 768 (16:9). The cheapest 1080p option is Wan 3.0 Video Prime at $0.20 / s.
Is video billed per second or per clip?
Per second of output for most models. Seedance 2.0 is billed per clip instead, at the same price for any supported length, which makes it the cheaper choice for longer clips. Generation time and queueing are never billed.
Do failed video generations cost anything?
No. A task that fails or is rejected upstream is refunded to your balance automatically, and every charge is itemized in your usage log.
Is there a subscription or minimum spend?
No. You top up a prepaid balance by card from $1 and pay per request at the rates on this page. Unused balance doesn't expire.
Generate your first clip
Create an account, top up from $1 and call any model above, or try one in the text-to-video playground first.