Reduce silence

Overview

The reduce_silence parameter shortens the pauses between words in the generated speech.

• false is the default and keeps the model's natural pacing.
• true shortens the pauses, so the delivery sounds tighter.

reduce_silence is separate from speech speed. speed changes how fast the words themselves are spoken, while reduce_silence only affects the gaps between them.

Model support

Not every model supports silence reduction. Set reduce_silence only on models where supports_silence_reduction is true in the List models response.

Enabling it on any other model returns an invalid_request error.

Usage

reduce_silence is supported by both the real-time WebSocket API and the REST API.

In the WebSocket configuration message:

{
  "api_key": "<SONIOX_API_KEY|SONIOX_TEMPORARY_API_KEY>",
  "model": "tts-rt-v2",
  "language": "en",
  "voice": "Adrian",
  "audio_format": "wav",
  "reduce_silence": true,
  "stream_id": "stream-001"
}

In the REST request body:

{
  "model": "tts-rt-v2",
  "language": "en",
  "voice": "Adrian",
  "audio_format": "wav",
  "text": "Hello from Soniox Text-to-Speech.",
  "reduce_silence": true
}