Skip to main content
POST
For editing, mark exactly one video as reference_type=edit_source; auxiliary videos, images and audios use asset. All input videos together allow up to 3 clips, 2–15 seconds each and at most 15 seconds in total. Identify the target in the prompt using its position in the video array. Editing supports an explicit duration of 4–15 seconds and an aspect_ratio; neither must match the source. imini does not expose smart duration -1. Existing credit rules remain unchanged.

Authorizations

​
Authorization
string
header
required

Pass Authorization: Bearer <YOUR_API_KEY> in the request header. Go to the API Keys management page to create and manage API Keys.

Body

application/json
​
model
string
required

Fixed value doubao/seedance-2.0.

​
prompt
string

The prompt. Supports Chinese, English, Japanese, Indonesian, Spanish, and Portuguese.

We recommend ≤ 500 Chinese characters or ≤ 1000 English words; overly long prompts may cause the model to ignore some details.

Example:

"First-person POV throughout. At dawn, pick a red apple; shake tea in a shaker; camera zooms in on the layered fruit-tea texture."

​
resolution
enum<string>
default:720P

Output resolution. Defaults to 720P.

Available options:
480P,
720P
​
aspect_ratio
enum<string>
default:adaptive

Output aspect ratio. Defaults to adaptive (auto-adapted by the model based on the input assets or prompt).

Available options:
16:9,
4:3,
1:1,
3:4,
9:16,
21:9,
adaptive
​
duration
integer
default:5

Video duration in seconds. Defaults to 5; integer in the range 4 ~ 15.

Required range: 4 <= x <= 15
​
generate_audio
boolean
default:true

Whether to generate audio in sync with the video. Defaults to true. Generated audio is mono.

​
reference_images
object[]

Reference image list. Up to 9 images. The input mode is determined by the reference_type combination:

  • Image-to-video (first frame): 1 first_frame
  • Image-to-video (first + last frame): 1 first_frame + 1 last_frame
  • Multimodal reference-to-video: 1–9 with asset; may be combined with reference_videos / reference_audios

Image requirements: formats jpeg / png / webp / bmp / tiff / gif; ≤ 30 MB per image; total request body ≤ 64 MB (avoid base64 for large files); width/height 300–6000 px; aspect ratio between 1:2.5 and 2.5:1.

Maximum array length: 9
​
reference_videos
object[]

Reference video list. Up to 3 clips; total duration across all videos ≤ 15 seconds; public URL only.

Video requirements: formats mp4 / mov; resolution 480P / 720P; 2–15 seconds per clip, ≤ 50 MB each; width/height 300–6000 px; aspect ratio between 1:2.5 and 2.5:1; pixel-product range [640×640, 834×1112]; FPS 24–60.

Maximum array length: 3
​
reference_audios
object[]

Reference audio list. Up to 3 clips; total duration across all audios ≤ 15 seconds. Cannot be used alone — must be accompanied by at least one reference_image or reference_video.

Audio requirements: formats wav / mp3; 2–15 seconds per clip, ≤ 15 MB each; total request body ≤ 64 MB (avoid base64 for large files).

Maximum array length: 3
​
extra_params
object

Supports Seedance's official extension parameters. All optional:

  • seed — integer in [-1, 2^32-1]. -1 or omitted uses a random seed; the same seed yields similar (but not guaranteed identical) outputs.
  • watermark — boolean, defaults to false. Whether to add a watermark to the generated video.
  • return_last_frame — boolean, defaults to false. When enabled, you can retrieve the generated video's last-frame image (PNG, no watermark, same dimensions as the video) via the query endpoint, useful for chaining multiple continuous video segments.

Response

Submitted successfully (asynchronous mode)

​
task_id
string

Unique task ID, used for subsequent polling

Example:

"task_2044275355061448704"

​
model
string

The model ID used for this task

Example:

"doubao/seedance-2.0"

​
created_at
string

Task creation time (ISO 8601)

Example:

"2026-04-15T04:43:20.233Z"

​
request_id
string

Unique request ID

Example:

"67e2e394-cf6f-4998-b293-ebf20d14dc26"