RelayDanceDocs / Core Concepts / How It Works
How RelayDance Works
RelayDance is a gateway between your application and the upstream model providers. You integrate one API and one key; the gateway forwards each request to the provider that serves the model you asked for.
#Request flow
- Authentication: your API key is validated and your balance checked.
- Routing: the request is forwarded to the upstream provider based on the
modelfield. - Response: the provider's output is returned to you in a consistent format.
#Synchronous vs asynchronous
Synchronous (image)
Image generation returns its result in the HTTP response, no polling required. RelayDance serves video and image generation only, there are no chat or LLM models.
Asynchronous (video)
Video rendering takes minutes, so the API never holds the connection open. POST /v1/video/generations returns a task_id immediately; you poll GET /v1/video/generations/{task_id}, or set metadata.callback_url and receive the final task state as a webhook.
| Status | Meaning |
|---|---|
submitted | Task accepted and queued |
running | Generation in progress |
succeeded | Done: the url field contains the clip |
failed | Generation failed and is not billed |
#Upstream providers
Video and image models from leading AI labs:
| Provider |
|---|
ByteDance (Seedance) |
HappyHorse |
OpenAI (gpt-image-2) |
#Billing
Usage is metered per request:
- Image generation: billed per request; image output is free, you pay only for input
- Video generation: billed per clip; price varies with model, resolution and duration
Failed requests are not billed: you only pay for output that is successfully delivered.