Multi reference image fusion for video generation

RelayDance supports multi reference image fusion by placing source media in metadata.content[] and citing each item in the prompt as @image1 to @imageN. A single request accepts up to 9 reference images, 3 reference videos and 3 audio tracks, with clips up to 15 seconds. You submit the task via POST /v1/video/generations on an OpenAI-compatible API, then poll or receive a webhook for the finished video url.

How reference media fusion works

Reference media are attached in metadata.content[] and referenced by name in the prompt text. Each image, video or audio entry maps to a token (@image1 through @imageN) that you cite inside the prompt to instruct the model where to apply the source. A single request supports up to 9 reference images, 3 reference videos and 3 audio tracks. Generated clips run up to 15 seconds. You define the model, prompt, and seconds, plus metadata fields such as ratio, resolution, generate_audio and callback_url. See relaydance.com/docs for the full field list and payload structure used to combine multiple references in one generation call.

Submitting and retrieving a fusion task

Submit a multi reference job with a single POST request, then read the result by polling or webhook. Send POST /v1/video/generations with your model, prompt, seconds and metadata (including content[]). Poll GET /v1/video/generations/{task_id} until status is succeeded or failed; the succeeded response contains the video url. Alternatively, set metadata.callback_url so the final state is POSTed to your server.

1. Attach references to metadata.content[] (up to 9 images, 3 videos, 3 audio tracks).
2. Cite each item in the prompt as @image1 to @imageN.
3. POST to /v1/video/generations with model, prompt and seconds.
4. Poll /v1/video/generations/{task_id} or wait for the callback.
5. Read the video url from the succeeded result.

OpenAI-compatible setup and authentication

You use fusion features through the same OpenAI SDK by changing the base URL. 据 relaydance.com/docs 官方文档「Set base_url to https://relaydance.com/v1 and keep your OpenAI SDK」, so no new client library is required. Authenticate with Authorization: Bearer YOUR_API_KEY; create keys at relaydance.com/console. Billing is pay-as-you-go per generated video, and 据 relaydance.com 官方文档「Failed or errored requests are never billed」. This provides direct OpenAI-compatible access to Seedance without a BytePlus enterprise account or KYC. Payments accept USDT and Stripe card. Reference media, prompt tokens and metadata are all passed within the standard request body of the video generations endpoint.

Model rates for fusion output

Prices depend on the model and resolution you select for the fused output. The following per-model rates apply, published at relaydance.com/models:

For image work with gpt-image-2, image output is free and only input is billed; image-to-image starts from about CNY 0.035, and 4K and 1K output cost the same. Live rates remain on the models page.

FAQ

How many reference images can I use per request? Up to 9 reference images, along with 3 reference videos and 3 audio tracks, in a single request.

How do I point a prompt at a specific reference? Cite each item in metadata.content[] using the tokens @image1 to @imageN inside your prompt text.

Am I charged if a fusion request fails? No. According to the relaydance.com docs, failed or errored requests are never billed.