How to mix reference videos and audio in one request: multi reference beyond images

To mix reference videos and audio in a single RelayDance request, place all reference media in the metadata.content[] array when you POST to /v1/video/generations, then cite each item in your prompt as @image1 to @imageN. One request supports up to 9 reference images, 3 reference videos, and 3 audio tracks, with output clips up to 15 seconds. Set model, prompt, and seconds, and poll until the task reports succeeded or failed.

What reference media a single request accepts

A single RelayDance request accepts up to 9 reference images, 3 reference videos, and 3 audio tracks, with generated clips up to 15 seconds long. All of these go into the metadata.content[] array alongside the core fields model, prompt, and seconds. You can also set other metadata such as ratio, resolution, generate_audio, and callback_url. Reference items are cited inside the prompt text using tokens from @image1 through @imageN, which tells the model how each piece of media relates to your instruction. Live per-model rates are listed at relaydance.com/models, so confirm pricing for your chosen model before submitting.

How to submit and retrieve a multi reference task

Submit a multi reference task by sending a POST to /v1/video/generations, then poll a GET endpoint until the task completes. Follow these numbered steps:

1. Create an API key at relaydance.com/console and send Authorization: Bearer YOUR_API_KEY.
2. POST to /v1/video/generations with model, prompt, seconds, and your references in metadata.content[].
3. Reference each media item in the prompt as @image1 to @imageN.
4. Poll GET /v1/video/generations/{task_id} until status is succeeded or failed.
5. Read the video URL from the result, or use webhook mode by setting metadata.callback_url so the final state is POSTed to your server.

According to the official relaydance.com docs, 「Failed or errored requests are never billed」, so polling errors do not add cost.

How billing works for multi reference generations

Billing is pay-as-you-go per generated video, and failed or errored requests are never billed. Rates vary by model and resolution: Seedance 2.0 720p costs about $0.190 per second, Seedance 2.0 1080p costs about $0.470 per second, Seedance Fast costs about $0.152 per second, and Seedance native 4K costs about $4.90 per 5-second clip (source relaydance.com/models). The number of reference items you attach does not itself appear as a separate line in these facts; cost is tied to the generated video.

How to connect using an OpenAI-compatible SDK

Connect by pointing your existing OpenAI SDK at the RelayDance base URL rather than rewriting your client. According to the official relaydance.com/docs, 「Set base_url to https://relaydance.com/v1 and keep your OpenAI SDK」. The protocol is OpenAI-compatible, so your request structure for /v1/video/generations and /v1/images/generations stays familiar. Authentication uses Authorization: Bearer YOUR_API_KEY, with keys created at the console. Payments are accepted via USDT and Stripe card. This provides direct access to Seedance without a BytePlus enterprise account or KYC. For current model availability and rates, see relaydance.com/models and the full reference at relaydance.com/docs.

FAQ

How many reference videos and audio tracks can one request include? One request can include up to 3 reference videos and 3 audio tracks, in addition to up to 9 reference images, with output clips up to 15 seconds.

How do I reference each media item in the prompt? Place each item in metadata.content[] and cite it in the prompt text as @image1 through @imageN.

Am I charged if a generation fails? No. According to the official relaydance.com docs, 「Failed or errored requests are never billed」.