Skip to content

Voice

Voice covers everything related to talking to NEXUS and having it talk back: how you start listening, which devices it uses, how speech is recognized, which provider reads responses aloud, an optional wake phrase, which commands ask for confirmation before running, and whether transcripts are kept on disk.

Voice is off by default, end to end. Nothing in this section listens, speaks, or writes a transcript until you turn on Enable voice control below, and each capability it depends on has its own further toggle on top of that.

Voice Control

LabelWhat it doesDefault
Enable voice controlMaster switch for voice control.Off
ActivationHow you trigger listening: Click to record, Keyboard shortcut, Push to talk (hold), Wake phrase, or Text only (no microphone).Keyboard shortcut
HUD styleCinematic or Compact voice overlay.Cinematic

Text only (no microphone) is worth calling out on its own: it lets you use voice commands' text-driven parts without ever opening the microphone. See Keyboard Shortcuts for the default shortcuts that start and stop a voice command once this section is on.

Audio Devices

LabelWhat it doesDefault
MicrophoneInput device used for voice.System default
Speak responses throughOutput device used for spoken responses.System default

Recognition

LabelWhat it doesDefault
Voice languageLanguage used for recognition.System default
Prefer on-device recognitionUses on-device recognition when available.On
Allow server-based recognitionFalls back to server-based recognition when needed.Off
Silence auto-stopHow long a silence before recording stops automatically.1.6 s (range 0.6–4.0 s)

On-device recognition is preferred by default; server-based recognition is an opt-in fallback, not a silent default, so nothing you say leaves your machine for recognition purposes unless you turn that on yourself.

Speech Provider

LabelWhat it doesDefault
ProviderWhich text-to-speech provider reads responses aloud: the built-in on-device voice, a self-hosted server, or a cloud provider you configure.Built-in
Test voicePlays a short sample with the current settings.

The built-in provider needs nothing further configured and keeps speech synthesis on-device. Choosing a self-hosted or cloud provider opens the connection details that provider needs; Test voice is the fastest way to confirm a provider is actually reachable before relying on it mid-session.

Responses

LabelWhat it doesDefault
Speak responsesReads responses aloud.Off
VoiceWhich system voice to use.System default
RateSpeaking rate.0.5 (range 0.35–0.7)
PitchSpeaking pitch.1.0 (range 0.5–2.0)

Dictation

LabelWhat it doesDefault
Open System Dictation…Opens macOS's own dictation settings.

Wake Phrase

LabelWhat it doesDefault
Enable wake phraseStarts voice control by saying a wake phrase instead of a shortcut.Off

The wake phrase is detected entirely on-device — enabling it doesn't send continuous audio anywhere, and it only applies when Activation above is set to Wake phrase.

Safety

LabelWhat it doesDefault
Confirm commandsWhich risk level of a voice-issued command requires confirmation before running: "Caution and above" or "Dangerous only".Caution and above

Whichever level you choose, a command NEXUS considers dangerous is always previewed before running — this setting narrows or widens what also gets a confirmation prompt below that top tier, it doesn't let a dangerous command skip one.

Privacy

LabelWhat it doesDefault
Store transcripts on diskSaves voice transcripts to disk.Off

Audio itself is never stored regardless of this setting — it controls only the text transcript of what was recognized, and that transcript isn't written anywhere unless you turn this on.

Where to go next

Built with purpose.