API de SeeDance 2.0 Mini

What is SeeDance 2.0 Mini?

The SeeDance 2.0 Mini API gives you the cost-optimized model in ByteDance's SeeDance 2.0 line. ByteDance released it in June 2026 as the third 2.0 model, next to Pro and Fast. All three share the same feature set. They differ in the balance between output quality, speed, and cost, and Mini sits at the cost end.

On UnificAlly the model ID is bytedance/seedance-2.0-mini. It covers three request types: text-to-video, first and last frame, and omni-reference. Videos run 4 to 15 seconds at 480p or 720p, with synchronized audio on by default. Billing is per second of output, from $0.022 per second.

So you get the full 2.0 control set at a lower cost per video. That makes Mini the model to run while a prompt, a reference set, or a shot list is still changing.

Key features of SeeDance 2.0 Mini

What SeeDance 2.0 Mini is best for

• Prompt testing

  Run many short 480p versions of one idea, compare them side by side, and keep only the prompts that work.

• Reference set checks

  Try different mixes of character images, motion videos, and audio cues before you commit a final render to them.

• High-volume social output

  Batch 9:16 and 1:1 videos for feeds and ads where the count of videos matters more than peak detail.

• Storyboard animatics

  Turn first and last frame pairs into moving shots so a team can review timing and camera before final renders.

• Product and listing videos

  Animate catalog shots with an omni-reference product image across many SKUs, each run at the same settings.

• Delivery files on a budget

  Render picked shots at 720p on Mini when the result already meets the brief.

SeeDance 2.0 Mini use cases

Build a draft loop for an ad tool: each user prompt renders at 480p for 4 seconds, and only the version the user picks gets re-run at 720p with the same seed. Feed a batch job a list of product photos and one motion reference video, so every SKU gets a matching 9:16 video. Give a storyboard app first and last frame support, where artists upload two key frames and get a moving shot back. Use omni-reference to test which character sheet and music cue suit a scene before moving the final pass to Pro.

SeeDance 2.0 Mini limitations

• Aspect ratio is only set in text-to-video. Frame and omni-reference requests take their shape from the inputs.
• Reference videos must be mp4 or mov, 480p to 1080p, 2 to 15.2 seconds each, and 15.2 seconds combined. Their input seconds are billed on top of the output.
• Audio references need at least one image or video reference in the same call, and cap at 15 seconds combined.
• Output stops at 15 seconds per task. Longer scenes need several tasks joined in the edit.
• Output tops out at 720p. For 1080p or 4K, send the same request to Pro.
• Pro produces the highest quality in the 2.0 line. For hero shots where detail and motion quality decide the result, test the same request on Pro.

When to use SeeDance 2.0 Mini

Start with Mini when the work is exploratory or high-volume: drafts, prompt tests, reference experiments, and batches where cost per video sets the budget. Move to SeeDance 2.0 Fast when turnaround matters more and the resolution range is the same. Move to SeeDance 2.0 Pro for final hero renders that need the best quality the line produces, or 1080p and 4K output. Because the three models accept the same fields, a team can draft on Mini, lock the seed and inputs, and re-run only the chosen version on Pro without rewriting a request.

SeeDance 2.0 Mini API examples

Call SeeDance 2.0 Mini from any language by POSTing to /v1/tasks. Full parameter docs live at docs.unifically.com/models/video/bytedance/seedance-2.0-mini.

curl -X POST https://api.unifically.com/v1/tasks \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -d '{
    "model": "bytedance/seedance-2.0-mini",
    "input": {
      "prompt": "A barista pours latte art in a sunlit cafe, slow push-in, soft morning light",
      "aspect_ratio": "9:16",
      "duration": 5,
      "resolution": "480p",
      "seed": 42
    }
  }'

Successful submission returns a task_id. Poll GET /v1/tasks/<task_id> or set a callback_url on the request to receive the finished result.