Introduction

RawFrame puts image and video generation models from several providers behind a single REST API. You send the same request to every model, pay from one prepaid credit balance, and get your files back from our storage, whichever provider made them.

How it worksCopy Anchor Link

1. Create a generation. POST /v1/generations with a model, a prompt and optional params and references. The request is checked against the model's capabilities, its maximum price is reserved from your credits, and it's queued. The response comes back immediately with status queued.
2. Wait for it. Poll GET /v1/generations/{id}, stream status over Server-Sent Events, or receive a webhook. Video usually takes one to several minutes; images take seconds.
3. Download the outputs. A succeeded generation has signed outputs[].url links to our storage, valid for an hour. Fetch the generation again any time for fresh links while the files are retained.

You're charged for what was produced, never more than the reserved estimate. Failed and canceled generations aren't charged. See Pricing and credits.

ConventionsCopy Anchor Link

• Base URL: https://api.rawframe.dev, versioned under /v1 (Versioning).
• Authentication: Authorization: Bearer <API key> (Authentication).
• JSON in and out. Field names are snake_case. Times are ISO 8601 in UTC.
• Money is integer micro-USD: 1 USD = 1,000,000. charge_micros: 40000 is $0.04.
• Ids have a type prefix: gen_ generations, upl_ uploads, ast_ output files, whe_ webhook endpoints.
• Errors share one shape with a stable code (Errors).