Seedance 2.0 or 2.5: choose by duration, resolution, and budget

Choosing a video model starts with the requirements of the request you can actually submit. SeedanceLink currently publishes two model IDs: Seedance 2.0 and Seedance 2.5. Both use the same task creation, status, and download flow. The concrete differences shown here are their permitted request durations and published retail prices, not an invented quality score or a guarantee that one will finish faster.

Use the following table as a request-planning reference. It describes the current SeedanceLink offering, rather than every product carrying the Seedance name. A feature listed on another service is not automatically part of this API. In particular, this guide does not promise 4K output, multiple videos in one task, editing, or advanced reference-material workflows that have not been included in this walkthrough.

Start with the supported ranges

A request from four to fifteen seconds can fit either model’s duration range. A request from sixteen to thirty seconds requires Seedance 2.5 in this catalog. There is no useful comparison between a supported request and one that the selected model cannot accept: choose a valid model first, then decide which resolution and price fit your job.

For a short integration test, use a common configuration such as four seconds, 480p, and a simple text prompt. That keeps the initial request inside both ranges and separates basic authentication or request-shape problems from decisions about longer output. Use the exact ID in the table; “Seedance 2.5” is a display name, while seedance-2.5-933 is the identifier sent in JSON.

Budget for the requested resolution

Both models publish 480p, 720p, and 1080p price tiers. At 480p, four requested seconds are USD $0.36 for Seedance 2.0 or USD $0.52 for Seedance 2.5 under standard retail rates. These are estimates from the same catalog used on the model page, not claims about supplier cost or a discount against another company’s price.

Resolution and duration both affect the quote. Test the model page’s estimator before choosing a longer, higher-resolution task. Your API key may have an applicable group multiplier, so the task receipt is the authoritative quote for that request. A requested resolution is a billing tier; do not use it to assume an exact media pixel height without inspecting the downloaded file.

Defaults matter when you migrate a sample between models. Omitting duration selects five seconds; omitting resolution selects 720p. Set these values explicitly when you compare costs. Otherwise two prompts that look similar in your code may reserve different amounts because the request settings are different.

Keep the integration unchanged

You can keep the create → check → download workflow and change the model field to choose the other model. Validate the duration against the newly selected model before you submit. If an application allows thirty seconds with Seedance 2.5 and the user switches to 2.0, adjust the duration or show a validation message; do not silently send an unsupported thirty-second task.

API_KEY='YOUR_API_KEY'
API_BASE='https://seedancelink.io'

curl --fail-with-body --request POST "$API_BASE/api/v3/contents/generations/tasks" \
  --header "Authorization: Bearer $API_KEY" \
  --header "Content-Type: application/json" \
  --data '{
  "model": "seedance-2.5-933",
  "content": [
    {
      "type": "text",
      "text": "A paper boat on a calm lake at sunrise."
    }
  ],
  "duration": 4,
  "resolution": "480p",
  "ratio": "16:9"
}'

Store the chosen model, resolution, duration, local task ID, and receipt together. Those fields let you explain the quote and later find the finished task without guessing which model was selected. Poll the saved task rather than creating a new task to “refresh” it. The same ownership and uncertain-submission rules apply to both models.

Model-selection questions