WebSocket improvements: reliability & custom timeout

We’ve made updates to our WebSocket API that make long audio generations more stable. These changes fix issues where voices could become robotic or fade during longer sessions, ensuring consistent quality throughout. Below, we've shared how a voice changes between the beginning and end of a long generation:

Before

After

We've also added a custom inactivity timeout for WebSocket connections, with a new maximum of 180 seconds. The default remains at 20 seconds but can be extended to better suit your needs.

To adjust this, simply add inactivity_timeout as a query parameter in the WebSocket endpoint:

wss://api.elevenlabs.io/v1/text-to-speech/{voice_id}/stream-input?model_id=eleven_turbo_v2&inactivity_timeout=180