AI Video Duration and Resolution Limits by Model and Their Cost

Duration and resolution shape a video request on MakerNice more than any other fields. Both are validated against the model you pick, and both feed into the price. Seedance 2.5 (ByteDance) and Wan 3.0 (Alibaba) allow up to 30 seconds, most other models stop at 15, and resolution options range from a single fixed tier to 4k.

What durations and resolutions does each video model allow?

duration is an integer number of seconds and must fall inside the model range. If you omit it, the default of 5 seconds is used for every model. If you omit resolution, the model default in the last column applies.

Which models take no resolution field?

Seedance 2.0 Fast and all Kling 3.0 models are priced at one fixed tier, so the request must leave resolution out; sending it is rejected. The tier is set by the model you choose: 720p for Seedance 2.0 Fast, 1080p for Kling 3.0 Pro, 720p for Kling 3.0 Standard and 4k for Kling 3.0 4K. Seedance 2.0 itself, unlike its Fast variant, accepts 720p or 1080p.

How does duration change the price of a video?

Video is priced per second: for a fixed model, mode and resolution, cost equals the unit price times the duration, so doubling the seconds doubles the price. In GET /v1/pricing, which needs no API key, every video row has its unit set to second.

Does a higher resolution always cost more?

Resolution does not scale cost smoothly. It selects a different row in the price list, and how far apart those rows are differs from model to model. Read the rows from GET /v1/pricing instead of assuming a ratio, and confirm the exact amount with a quote.

Should you always pick the maximum duration?

No. Every extra second is billed, and a clip that is longer than the idea needs usually feels slower rather than richer. Choose the shortest duration that carries the action in your prompt, and use the long ranges, such as the 30 seconds that Wan 3.0 and Seedance 2.5 allow, only when the shot genuinely needs them.

What is a good draft-then-final workflow?

Prompts rarely land on the first try, so test small and finish at full quality. Draft at the shortest sensible duration and the lowest resolution the model offers. Once the prompt and composition are right, re-quote with the final duration and resolution and generate the real clip.

{
  "model": "wan-3.0",
  "mode": "t2v",
  "prompt": "A lighthouse on a cliff at sunrise, waves rolling in, slow aerial pull-back",
  "duration": 4,
  "resolution": "480p",
  "aspect_ratio": "16:9"
}

Changing duration or resolution means a new quote, because both affect the price; the final run is a separate quote and a separate create with its own idempotency key.

For the cost side, check the Pricing page or request a quote with your final settings; for the creative side, read the text-to-video prompt tips post.