Soniox provides a WebSocket-based API for real-time and low-latency speech recognition using Soniox Speech Service. This enables you to integrate with Soniox from any programming language or framework that supports WebSockets (e.g. Swift).

Soniox Web Library (a Javascript client library) uses WebSocket API to integrate with Soniox Speech Service. See the Soniox Web Voice GitHub repository for the implementation details.

The Soniox WebSocket endpoint address is wss://api.soniox.com/transcribe-websocket.

If you are going to integrate with Soniox via WebSocket API, we recommend taking a look at the following examples:

Temporary API Keys#

To enable using the Soniox WebSocket API from a web browser without exposing your regular Soniox API key to web clients, you can create temporary API keys. Each created temporary API key is valid only for a single transcription using WebSocket and only for a limited time after creation. Refer to: