Video generation guide
Choose Seedance or another video model, create a task, and download the result.
Choose a model and check its supported inputs
Query the video catalog with the API key you plan to use. Use the returned model ID and parameter schema to check text, first-frame, and multimodal input support.
Seedance 2.0 series
Choose a variant available in your catalog. Check its schema for supported roles and combinations before sending a reference image, video, or audio clip.
Seedance 2.5
First-frame and first/last-frame output follows the first frame’s aspect ratio. Omit ratio on the Volcengine-compatible path; use the selected variant’s schema on other paths.
Other video models
Use the unified API for models such as Wan or MiniMax H3 when they appear in your catalog. Follow the model schema for duration, resolution, and assets.
Create a task and retrieve the video
Use the unified API for a new integration or for other video models. Check the catalog and schema with your API key for model and input availability.
- Create
POST https://intertoken.ai/v1/contents/generations/tasks- Query the original task
GET https://intertoken.ai/v1/contents/generations/tasks/{id}- Read on success
output[].url
Migrate existing Volcengine Ark Seedance code by switching to the TokenPortal address, API key, and model ID. This path is for Seedance only.
- Create
POST https://intertoken.ai/api/v3/contents/generations/tasks- Query the original task
GET https://intertoken.ai/api/v3/contents/generations/tasks/{id}- Read on success
content.video_url
Store the returned task ID and query it with the same key and API path. If the create result is uncertain, check the retry instructions for that path before attempting another submission.
Choose an input
Adapt the request for your API path and model. Replace MODEL_ID with the catalog ID and the example URL with your own asset URL.
Describe the scene in a prompt and set generation parameters according to the model schema.
{
"model": "MODEL_ID",
"content": [
{
"type": "text",
"text": "A slow camera pan over a city at dusk."
}
]
}Use an image to set the opening shot with the first_frame role.
{
"model": "MODEL_ID",
"content": [
{
"type": "text",
"text": "A slow camera push from this first frame."
},
{
"type": "image_url",
"role": "first_frame",
"image_url": {
"url": "https://example.com/first-frame.jpg"
}
}
]
}Generate a video from a first-frame imageCheck the model schema for supported reference types and combinations. Do not mix multimodal references with first-frame or first/last-frame modes in one request.
{
"model": "MODEL_ID",
"content": [
{
"type": "text",
"text": "Use the image as the visual reference."
},
{
"type": "image_url",
"role": "reference_image",
"image_url": {
"url": "https://example.com/reference.jpg"
}
}
]
}Register and reference assetsFirst frames and reference assets
Prepare inputs according to the model schema. Use a public URL when supported. If an asset reference is needed, register the asset, wait until it is ready, and use the returned ref.