Article
3 min read
UllrAI Lab

Seedance 2.5 Draft Mode: 480p Previews, 1080p Finals and Cost

Use Seedance 2.5 Draft mode without confusing BytePlus and Runway APIs. Understand the seven-day window, separate charges, and when previews actually save money.

Seedance 2.5 Draft mode lets you review a 480p preview before generating a 1080p final. Both stages are charged. Its value is avoiding full-resolution generation for ideas you will reject, not getting a free preview or a discounted final.

BytePlus now documents the native workflow

The current ModelArk tutorial documents dreamina-seedance-2-5-260628, normal output at 480p, 720p and 1080p, and a two-stage Draft workflow. This supersedes older documentation that limited the API to 720p. It does not establish 4K API support.

  1. Create a task with draft: true and resolution: "480p".
  2. Wait for succeeded, inspect the output, and save the Draft task ID and created_at time.
  3. Use the documented Draft-task reference to create the final within seven days of that timestamp.
  4. Keep the same model, omit draft or set it to false, and set final resolution to 1080p.
  5. Do not resend inherited prompt, references, duration, ratio, seed, audio, or task-type settings. The tutorial says these cause an error even when unchanged.

Use the current request example for the complete task-reference structure. An API field from another provider is not interchangeable with ModelArk's schema.

Runway exposes a separate API contract

Runway uses model ID seedance2_5. Its documentation also describes 480p previews and enhancement to 1080p, with a successful Draft task owned by the same account and a seven-day window. Use Runway's task reference and pricing, rather than sending its fields to BytePlus. Store the provider alongside every task ID.

How billing works

BytePlus bills the Draft as a normal 480p video and the final at its target resolution. In the final stage, the rate and input-video duration depend on the original inputs; the Draft output itself is not counted as a new video reference. Actual charges follow token usage, including applicable minimums for video input.

The pricing page's five-second, 16:9 examples without video input are $0.514 at 480p and $2.843 at 1080p. These are published estimates, not a fixed per-video price for every aspect ratio or request.

Using those examples only, three Draft candidates and one final cost about $4.385, compared with $8.529 for three direct 1080p attempts. If you need just one successful direct attempt, adding a Draft costs more: about $3.357 instead of $2.843. These calculations assume the same duration and a final you keep; they are not measured savings.

Review the right things before enhancing

Check composition, character placement, action timing, camera movement, and continuity. A small preview may hide hand, face, text, or audio defects, so review the final separately. Keep enough time for that final review before the seven-day window closes.

Record preview and final task IDs, provider, model, inputs, review decision, expiry, and the actual charges. A production budget should divide total spend by approved final seconds, not by the number of successful API calls.

For model access and broader constraints, see our Seedance 2.5 guide. For planning shots before generating, use the storyboard workflow.

Sources reviewed October 21, 2026: BytePlus Draft tutorial, BytePlus pricing, Runway model documentation.

Model and product names in this article are trademarks of their respective owners. AniKuku and UllrAI Lab are not affiliated with or endorsed by them.

Plan and produce your next AI video

Manage prompts, characters, storyboards, and video shots in one project, then compare models when you need to.