Models Overview

The Seedance 2.0 API is not a single model — it's a matrix of 6 models. You must specify exactly one of them in the model field of every request. There is no "automatic mode detection"; sending the wrong model ID will return model_access_denied or invalid_request.

The Model Matrix

Input typeStandard (best quality)Fast (quicker / cheaper)
Text onlyseedance-2.0-text-to-videoseedance-2.0-fast-text-to-video
1–2 imagesseedance-2.0-image-to-videoseedance-2.0-fast-image-to-video
Images + videos + audioseedance-2.0-reference-to-videoseedance-2.0-fast-reference-to-video

All six models share the same endpoint:

POST https://api.evolink.ai/v1/videos/generations

The only differences are the model field value and which input arrays (image_urls / video_urls / audio_urls) each model accepts.

How to Choose a Model

Two steps: pick the mode based on your inputs, then pick Standard vs Fast.

Step 1: Pick the mode by input

Text prompt only
    → text-to-video  (optional web_search parameter for up-to-date content)

1 reference image  (used as first frame)
2 reference images (used as first + last frames)
    → image-to-video

Multiple images, reference videos, or audio tracks  (up to 9 + 3 + 3)
    → reference-to-video

Step 2: Standard or Fast

DimensionStandardFast
Visual quality stabilityHigherAcceptable
Generation speedBaselineFaster
Per-second costBaselineLower
Recommended forFinal deliverables, ads, hero product videosPreviews, A/B testing, bulk short-form production

Tip: The same prompt can usually be switched between Standard and Fast with zero code changes — the parameter structure is identical, only the model field differs. Use Fast during prompt iteration, then swap to Standard for final delivery.

Capabilities Shared by All Models

Regardless of which model you pick, Seedance 2.0 supports:

  • Synchronized audio generation — generate_audio: true (default). Put dialogue inside straight double quotes to optimize speech synthesis, e.g. The man said: "Remember, never point at the moon."
  • Multiple aspect ratios — 16:9, 9:16, 1:1, 4:3, 3:4, 21:9, adaptive
  • Two quality tiers — 480p or 720p (1080p is not supported)
  • 4–15 second durations — any integer, default 5 seconds
  • Async task lifecycle — every request returns a task ID immediately; retrieve the result by polling or via callback_url webhook

Billing uses per-second pricing (billing_rule: "per_second") — longer duration values cost more for the same clip.

Exclusive Capability: model_params.web_search (text-to-video only)

seedance-2.0-text-to-video and seedance-2.0-fast-text-to-video support an additional model_params.web_search parameter. When enabled, the model autonomously decides whether to search the internet for up-to-date information (and is only billed when a search is actually triggered):

{
  "model": "seedance-2.0-text-to-video",
  "prompt": "An ad for the latest 2026 spring-edition electric sports car",
  "model_params": {
    "web_search": true
  }
}

Prompt Length Limit

All models enforce the same prompt limit:

  • Chinese: ≤ 500 characters
  • English: ≤ 1000 words

Going over is rejected outright. Longer prompts do not produce better results — focus on the subject, action, and cinematography.

Next Steps

Learn More