Voice Input

The mic control in the lower-left lets you dictate a prompt instead of typing. With the window focused, press the configured trigger key or click Listen to start recording; press again (or click Stop) to finish. The control reflects its state (LISTEN → LISTENING → WORKING), transcribes the audio, and drops the text into the input box. By default voice is enabled, the trigger key is F1, and the transcript is submitted automatically through the Ask path. Recording is toggle-based (not push-to-talk) and requires window focus; there is no global voice hotkey. Raw audio is kept in memory and never written to disk.

Voice behavior is configured under the gui.voice section of the config (see Configuration):

Setting	Default	Meaning
gui.voice.enabled	true	Master on/off for voice input
gui.voice.trigger_key	F1	Key that toggles recording
gui.voice.auto_submit	true	Submit the transcript automatically once ready
gui.voice.max_recording_duration	60s	Stop recording after this long
gui.voice.stt.backend / gui.voice.stt.model / gui.voice.stt.language	openai / gpt-4o-mini-transcribe / —	Speech-to-text backend, model, and language

The default OpenAI backend sends recorded audio to OpenAI for transcription and reuses your OpenAI auth. Voice is disabled while a response is in flight so it cannot collide with a running prompt.