Skip to main content
POST
Pricing: See vibetool.ai/pricing for current rates.
Seedance 2.5 Reference To Video is ByteDance’s latest video generation model. It produces 480p or 720p videos from text prompts, still images, or existing video references, and supports both editing and extending footage. Videos can run from 4 to 30 seconds and include native audio. Billing is per second of generated video.

Input Modalities

  • Text – a prompt describing the scene, motion, or edit you want
  • Image – a still image used as a visual reference
  • Video – an existing clip to extend or edit using the new prompt

Output Modality

  • Video – an MP4 file with audio. Resolution is 480p or 720p; duration is 4–30 seconds depending on request and model limits.

Authorization

string
required
All APIs require Bearer Token authentication. Format: Bearer YOUR_API_KEY

Request Body

string
required
The model identifier. Use "seedance-2.5-reference-to-video".
string
Text description of the video to generate, or the edit to apply to a reference image or video. Required when neither image_url nor video_url is provided.
string
A publicly accessible URL to a reference image (JPEG or PNG). Use this for image-to-video generation.
string
A publicly accessible URL to a reference video (MP4 or similar) for editing or extending existing footage.
number
Requested video duration in seconds, between 4 and 30. If omitted, the model chooses a duration within that range.
Provide at least one of prompt, image_url, or video_url.

Response Schema

string
required
Unique identifier for the generation task.
string
required
Type of object, always "video.generation".
string
required
The model used, e.g., "seedance-2.5-reference-to-video".
integer
required
Unix timestamp (seconds) of task creation.
string
required
Initial status: "pending" or "processing". Use the Get Video Status endpoint to poll for completion.
string
required
The task ID to query in the status endpoint.
Once the task is created, poll the Get Video Result endpoint at GET /v1/videos/status/{task_id} until status is completed, then use the returned video URL. The task_id from the response above is used in the URL path.

Authorizations

Authorization
string
header
required

Bearer authentication header of the form Bearer <token>, where <token> is your auth token.

Body

application/json
model
enum<string>
required

Use seedance-2.5-reference-to-video.

Available options:
seedance-2.5-reference-to-video
prompt
string
required

Text description of the video to generate

image_urls
string<uri>[]

Reference images

Maximum array length: 30
video_urls
string<uri>[]

Reference video clips

Maximum array length: 10
audio_url
string<uri>

Reference audio

duration
integer
default:5

Duration in seconds (4-30)

Required range: 4 <= x <= 30
quality
enum<string>
default:720p

Output resolution tier

Available options:
480p,
720p

Response

Task created

id
string

Unique identifier for the generation task

object
string
Example:

"video.generation"

model
string
created_at
integer

Unix timestamp (seconds)

status
enum<string>

Initial task status

Available options:
pending,
processing
task_id
string

Task ID for the status polling endpoint