Become a provider

What video orchestration backends need to implement to integrate with Pipevideo as a provider.

This guide is for teams building a video orchestration backend that want to be listed as a provider on Pipevideo. End users integrate only with the Pipevideo API — your service is called by Pipevideo, not by customer applications directly.

What a provider does

A provider fulfills the heavy work of video generation after Pipevideo accepts a customer request:

StageResponsibility
OrchestrationRun the customer's selected orchestration model to compose the video project from the prompt (and optional reference media)
RenderingProduce the final video using a supported engine (e.g. HyperFrames, Lottie)
ArtifactsOptionally store secondary outputs alongside the video (e.g. source animation JSON, transcript logs)
Usage reportingReturn input/output token counts (and optional cost) so Pipevideo can bill the customer
Completion callbacksNotify Pipevideo when a job finishes so the gateway can update status and fire customer webhooks

How you implement orchestration, sandboxing, and rendering is up to your platform. Pipevideo does not document provider internals in customer-facing docs.

Integration model

Generation is asynchronous:

  1. Pipevideo creates a generation record and forwards work to your API with a callbackUrl.
  2. Your service returns an external job id and processes the request.
  3. On completion or failure, you POST to the callback URL with a signed payload.
  4. Pipevideo maps the result to the customer's gen_... id and delivers webhooks if configured.

Your HTTP API should support at minimum:

  • Create generation — accept engine, orchestration model, prompt, dimensions, and callback URL; return an external id.
  • Status / poll (optional if callbacks are reliable) — return status, output URL, artifacts, and usage.
  • Model catalog — orchestration models your provider supports per engine, for listing in GET /v1/engines.

Pipevideo holds provider API keys and webhook secrets in its environment — customers never see them.

Provider contract (conceptual)

Each provider registers a slug (e.g. framerate) and declares:

  • supportedEngines — engine ids this provider can render
  • orchestrationModels — model ids and pricing metadata for the public catalog

When Pipevideo routes a request, it passes:

  • generationRequestId — correlates your job with the customer's generation
  • engine, model, prompt, width, height
  • callbackUrl — where to notify on completion

Your completion callback should include at least the external job id and terminal status. Pipevideo verifies callbacks with a shared secret (HMAC signature).

Catalog and routing

After integration, your provider appears in:

  • The Providers sidebar and provider routing catalog
  • GET /v1/engines — engines and orchestration models
  • Customer provider.order routing — e.g. { "order": ["your-slug"] }

Customers choose providers through Pipevideo only. Do not ask Pipevideo customers to call your API directly unless they are building a separate integration with your product.

Next steps

Contact the Pipevideo team with:

  • Provider name and proposed slug
  • Supported engines and orchestration models
  • Base URL for your provider API (staging and production)
  • Webhook signing approach