Overview
Running your custom speech model via the Blynt API.
Blynt builds a speech recognition engine tuned for your product and runs it as a dedicated deployment. You reach that deployment with a deployment id and an API key, both provided by Blynt.
There are two APIs, and they are alternatives. Which one you use is fixed by your deployment: a deployment is provisioned either for real-time streaming or for synchronous transcription, and the deployment id you were given determines which of the two it answers on. It is not a per-request choice. If you need both, you need one deployment of each - talk to the Blynt team.
The two APIs
Real-time API
Stream audio over a WebSocket and receive transcripts as the user speaks. For voice agents and live interfaces.
Sync API
POST a complete audio file over HTTP and get the transcript in the response. No session to manage.
| Real-time API | Sync API | |
|---|---|---|
| Transport | WebSocket (wss://) | HTTP POST |
| Provisioned for | audio that arrives as the user speaks | recordings that are already complete |
| Results | turn_partial while decoding, turn_ended at the end | one JSON response |
| Unit of work | one connection, many conversational turns | one request, one audio file |
| Languages | fr, en, pt, de, es, it | fr, en, pt, de, es, it |
| Contextual biasing | session-level and per turn | one context per request |
| Limits | the lifetime of the connection | 48 MiB and 480 seconds per request |
| Framework plugins | Pipecat, LiveKit | — |
Both take the same API key and the same contextual biasing shape, so what you learn on one side carries over to the other.
Calling the wrong API for a deployment fails. A deployment provisioned for Real-time answers 400 on POST /transcribe; a deployment provisioned for Synchronous transcription rejects the WebSocket connection. If you are unsure which one your deployment serves, ask the Blynt team.
Quickstarts
Stream a local audio file
Run a local audio file through the Real-time API with a copy-paste Python client.
Pipecat integration
Use Blynt as the STT service in a Pipecat voice-agent pipeline.
LiveKit integration
Use Blynt as the STT service in a LiveKit Agents session.
Transcribe a file
Send a complete recording to the Sync API with curl or Python.
Learn more
Terminology
Deployments, sessions, turns, and turn-taking modes explained.
Contextual biasing
Ground a session with what you know, and narrow a turn to what you expect.
Real-time API reference
The full schema for every client and server WebSocket message.
Sync API reference
The synchronous HTTP endpoint: request shape, response, and errors.