Skip to navigation

Flux TTS normalizes volume by default

As of Nov 8, 2026, Flux TTS (/v2/speak) levels the loudness of generated speech by default. If you already use Flux TTS, your audio may sound more uniform in volume, and some quieter voices will be louder. To keep the previous behavior, set normalize_volume=false.

  • What it does. Quiet output comes up, loud output comes down, and speech lands at a consistent level across voices and requests. A peak ceiling prevents clipping and distortion on playback devices, and it stays on even when you opt out.
  • Where it applies. Batch (REST) and streaming (WebSocket). On streaming, set it as a query parameter when you open the connection.
  • How to opt out. Batch: POST https://api.deepgram.com/v2/speak?model=flux-haley-en&normalize_volume=false. Streaming: wss://api.deepgram.com/v2/speak?model=flux-haley-en&normalize_volume=false.

See Volume normalization for when to turn it off, and the normalize_volume API reference for the parameter.