Skip to main content
You can use your own portrait image or video with Seedance in two ways. We recommend creating the Asset Group and Assets yourself. This lets you organize your portrait library, confirm each asset is ready, and reuse its Asset ID across video-generation requests. Alternatively, send an HTTPS media URL in your video-generation request. GENGEN creates or reuses an Asset Group, registers the media, waits for it to become ready, and then submits generation. Both methods below use dreamina-seedance-2-0-260128. You can also use dreamina-seedance-2-5-260628 with these examples. Model-specific media and output limits still apply; see Seedance 2.0 and Seedance 2.5. This method lets you prepare a portrait once and reuse its Asset ID in later generation requests. A Group organizes assets; each uploaded image or video receives its own Asset ID.

Step 1: Create an Asset Group

Call Create an asset group:
Request JSON:
Response excerpt:
Save Result.Id as your Group ID. If you already have a group owned by your user in the current workspace, reuse it instead of creating another. You can find existing groups with List asset groups. This create-group endpoint uses GroupType: "AIGC" for Virtual Portrait groups.

Step 2: Add a portrait image or video to the group

Call Create an asset:
Image request JSON — replace GroupId with the value from Step 1:
To add a video, set AssetType to Video and replace URL with the video URL. The endpoint, response format, and status-checking steps are the same for both media types. This example uses Moderation.Strategy: "Skip", as in GENGEN’s automatic portrait registration flow. It is available only where enabled. If the API rejects that option, contact support; omitting Moderation uses the standard moderation policy. Generation content checks still apply. Response excerpt:
Save Result.Id as your Asset ID. This response means preprocessing was accepted; it does not mean the asset is ready. Repeat this step for additional images or videos, saving each returned ID. No local file path, browser blob: URL, or Base64 upload is accepted by this flow.

Step 3: Wait for the asset to become Active

In most cases, asset preprocessing returns a result within 2 seconds. It can also fail; check the returned status and error details, and continue only when Result.Status is Active.
Images and videos depicting public figures or copyrighted characters are not allowed.
Call Get an asset. This is a POST request:
Request JSON — use the Asset ID from Step 2, without an asset:// prefix:
An asset that is ready returns:
Use the returned status to decide when to continue, rather than assuming a fixed processing time. If you uploaded multiple assets, wait for every asset you plan to use to become Active.

Step 4: Generate with the Asset URI

Prefix the Asset ID with asset:// and call Create a video task:
Request JSON:
Save the returned top-level id and follow the polling steps below. Requests using only existing Asset URIs do not need the automatic URL-registration stage. To reuse this portrait in another video, submit a new generation request with the same Asset URI; you do not need to recreate the group or re-upload the image while the asset remains available and accessible.

Method 2: Send an HTTPS URL

Step 1: Submit the portrait with your video request

Call:
Request JSON:
For example, save this JSON as request.json and send it with cURL:
You do not need to call the group or asset endpoints first. GENGEN automatically prepares the media in your API asset group, then starts generation when the assets are ready. This automatic flow applies to the public Seedance video task endpoint shown above.

Step 2: Save the returned video task ID

An accepted URL request returns HTTP 200 with a response like:
preparingAssets means GENGEN is still preparing the reference media. The returned id is the video task ID, not an Asset ID. Save the complete value, including its seedance20: or seedance25: prefix.

Step 3: Poll the same video task

Follow Poll the video task below. The task ID stays the same from asset preparation through video completion; you do not need to obtain or substitute the automatically created Asset IDs.
Do not repeat the generation POST to check progress. Each new POST creates a separate generation request and can incur another charge. Use GET with the returned task ID.

Poll the video task

Both methods finish with a video task ID. Use the same API key that created the video task to call Retrieve a video task:
Replace the entire sample ID with the id returned by your generation request. GET has no JSON body. Poll at a moderate interval, for example every five seconds, and continue while the task is pending or processing. A completed Seedance task returns a response like:
Read the video URL from outputs.videos[0]. If a compatible response reports succeeded, also treat that as success. Stop polling on failed, cancelled, or expired; inspect the returned error for the next action. An HTTP 200 response alone does not mean generation succeeded.

Use a portrait video instead of an image

For Method 1, create an asset with AssetType: "Video", wait for Active, and put its Asset URI in assets.referenceVideos. For Method 2, put the HTTPS video URL in that field. Refer to it as Video 1 in the prompt. For example, the manual video-asset request is:
Send it to POST /api/gengen/v1/assets, then poll POST /api/gengen/v1/assets/get with its returned Asset ID as described above, until its status is Active.