Overview
How OffRail exposes NSFW-capable generation models through a managed multimodal API.
OffRail fronts a focused catalogue of NSFW-capable generation models with one managed API. Your application addresses an OffRail model ID; the gateway handles eligible upstream access, request metering, and the delivery lifecycle for that modality.
Public release scope
The first public release is intentionally focused:
- image, video, chat, and speech generation routes;
- usage-based billing without a subscription requirement;
- managed upstream routing behind each public model ID;
- a live model catalogue as the source of truth for availability.
Self-service console access and API key provisioning are still in private preview. There is no separate demo route or public dashboard workflow in this release.
What stays consistent
Every public route uses the same core contract:
- Authenticate with an OffRail API key.
- Send the selected public model ID in the request.
- Let OffRail resolve an eligible upstream mapping.
- Receive a response or track a task using the documented delivery shape.
- Pay for measured usage in the route's native billing unit.
What changes by modality
| Modality | Primary input | Result | Lifecycle |
|---|---|---|---|
| Chat | Messages and controls | Text or structured response | Sync or streaming |
| Image | Prompt and optional reference | Image URL, base64, or task | Sync or task |
| Video | Prompt and optional references | Pollable video task | Async |
| Audio | Text and voice selection | Binary audio payload | Sync |
Model-specific sizes, durations, voices, and prices are exposed on the public catalogue and the relevant API reference page.
Managed routing
OffRail separates the model your application requests from the upstream provider that serves it. This lets the public model ID remain stable while eligible provider selection and supported fallback happen behind the gateway.
Fallback never changes the route contract. A video request remains an asynchronous video task; an audio request remains an audio response.
Usage-based billing
Billing follows the work performed by each route:
- chat is metered by input and output tokens;
- images are metered per generated image;
- video is metered per generated second;
- speech is metered by input characters.
See Pricing for the public billing model and each model detail page for its current starting rate.
Next step
Continue to the Quickstart when you have a private preview API key.