Volcano Engine reportedly opened developer access to Seedance 2.5’s 4K API on November 16, 2026. Separately, BytePlus ModelArk documents a Draft workflow in which a 480p preview and a final render at 1080p or 4K are billed as two tasks.
Budget for both the preview and the final render: Draft is a separate paid generation task.
BytePlus lists Seedance 2.5 output at 4K and 10-bit color. For a 16:9 frame, 4K is 3840 × 2160 pixels. Its Seedance 2.5 API documentation sets out the Draft process and model limits; those details describe BytePlus ModelArk.
How the Draft preview and final render work
On BytePlus ModelArk, a Draft request generates a 480p preview. If it meets the project’s needs, a second task uses the Draft task ID to render at 1080p or 4K. BytePlus bills the preview at its ordinary 480p rate and bills the final task at the selected resolution.
For the final request, use the same model ID and specify the target resolution. The service reuses the original prompt, references, duration, aspect ratio, seed, audio setting, and task type; sending those fields again can cause an error. Draft task IDs created on or after October 28, 2026, at 10:00 UTC+8 remain valid for 14 days. Seedance 2.5 Flash does not support Draft.
What BytePlus’s 4K cost estimates cover
BytePlus prices video generation by token rate and token consumption. Its estimates account for input and output duration, output dimensions, and frame rate, so a five-second output does not have one universal price.
| BytePlus ModelArk scenario | Input video | Output | Estimated cost |
| 4K generation | None | 16:9, five seconds | $4.568 |
| 4K generation | 2–30 seconds | 16:9, five seconds | $5.074–$19.732 |
These are BytePlus estimates for the stated scenarios. In Draft mode, the preview is an additional billable task; the figures above describe the final 4K output, not a combined Draft-and-final total. BytePlus’s pricing documentation explains the token-based estimates.
Which model supports 4K and Draft?
BytePlus lists Seedance 2.5 for 4K output and Draft. Seedance 2.5 Flash tops out at 720p and does not support Draft. Both models have a documented output duration of 4–30 seconds, while their reference-asset limits differ.
| Model | Maximum listed resolution | Draft | Maximum reference assets per request | Output duration |
| Seedance 2.5 | 4K, 10-bit | Yes | 30 images, 10 videos, and 10 audio clips | 4–30 seconds |
| Seedance 2.5 Flash | 720p, 8-bit | No | 10 images, 5 videos, and 5 audio clips | 4–30 seconds |
For BytePlus, the documented model ID is dreamina-seedance-2-5-260628; Volcano Engine’s documentation uses doubao-seedance-2-5-260628. The identifiers belong to separate platform APIs, as shown in the BytePlus and Volcano Engine guides.
BytePlus setup and 4K operating limits
BytePlus setup requires an account, an API key, and model activation. Its guide describes qualifying routes that include a resource pack with remaining quota, a balance above $30, or a dedicated AI Savings Plan of at least $30.
For 4K generation, BytePlus limits accounts to one concurrent task and 15 task-creation requests per minute. The API supports online inference; offline inference is not supported. Generated video URLs remain valid for 24 hours and can be downloaded up to 100 times.