Speech

Speech

Pick the microphone, transcription model, and voice output behavior.

Microphone device

Choose the input device Agent Zero listens to.

Speech-to-text model size

Larger models can hear more accurately, but need more time and memory.

Enable Kokoro TTS

Use higher-quality server-side speech instead of the browser voice.

chevron_rightAdvanced Settings

Language hint

Use a short language code such as en, fr, or it when automatic detection needs guidance.

Silence threshold

Lower values catch softer speech; higher values ignore more room noise.

End-of-speech delay

How long silence must last before Agent Zero treats your sentence as complete.

Microphone close delay

How long Agent Zero waits before closing the microphone after speech stops.