Skip to main content
1

Request access

Open models are on Hugging Face. For the API image, request access with your use case and languages. Approval is manual, not instant.
2

Local check (weights)

Then load the checkpoint named on the model page (ASR, BERT, or MT).
3

Get your API key

Once approved, you receive an API key with the image tag and runbook. Every /v1 route expects it as:
4

Confirm the box is up

Expected: { "status": "ok" }. This is a liveness probe, not a healthcare API.
5

Read Conventions, then an endpoint

Conventions covers language codes and audio format. Then pick one: Transcriptions, Speech, Translate, Embeddings.
6

Read the license

Confirm terms on Licenses before any redistribution or wrapping.