Skip to main content
Post

Flux TTS is generally available, and is now the default agent voice

Deepgram made Flux TTS generally available and the default voice provider in the Voice Agent API. A Settings message that omits agent.speak now uses flux-kit-en instead of aura-asteria-en. The catalog includes 36 English voices across seven accents.

Check your audio settings before relying on the new default. Flux accepts only linear16, mulaw, and alaw raw audio, with no container or bit rate. Sessions that omit agent.speak and request mp3, opus, flac, aac, or a container such as wav now receive INVALID_SETTINGS.

To keep the previous behavior, explicitly set agent.speak.provider.version to v1 and name an aura-* model.

Loading replies…