Voice settings

Choose local or cloud spoken replies, preview voices, and manage the controls that affect speech.

Open Settings → Agent → Agent voice to choose how Nock speaks its answers. Voice output is separate from microphone input: dictation can work while replies are silent, and spoken replies can be local even when the answer itself came from a hosted model.

Choose where replies are spoken

Under Where replies are spoken, choose On this computer or Cloud voice. The local choice synthesizes speech on the computer; cloud voice sends reply text to its configured speech route. The card shows whether a local voice download is needed, ready, or unavailable on this computer.

When On this computer is selected, preview one of the local voices. If the natural voice model is missing or the computer cannot keep up, Nock may use a simpler local voice or system voice. Local speech is currently intended for English; other languages may use the system voice. See on-device spoken replies.

With Cloud voice, choose a listed voice and speed. In Settings → Providers, an xAI key can be used for speech-to-text and cloud spoken replies; a Deepgram key is for speech-to-text. Your own key uses the provider’s account and terms. A key does not automatically grant every provider feature or account capability.

Preview and adjust speed

Choose the speech route.
Open the voice picker and press its preview button.
Select a voice and choose Slow, Normal, or Fast.
Ask a short question and compare the spoken line with the answer on screen.

Nock may speak only a short version while showing the full answer. Read the visible response for exact instructions, code, or longer details.

Stop or disable speech

Turn off Enable Agent voice to keep answers on screen without spoken playback. Starting a new voice request can interrupt the current speech. Interaction sounds and welcome music are separate settings under Settings → General.

The separate Speaks first control under Settings → Agent is off by default. If you enable proactive check-ins, you can choose how often they happen, which check-ins are allowed, and quiet hours (22:30–08:00 by default). You can say Not now to quiet check-ins temporarily or use Talk again to resume them. This does not change how ordinary answers are spoken. See Speaks first for the check-in triggers and controls. Reminder quiet hours are separate under Settings → Nudges; see Reminders.

Custom voices

Custom voice creation is available only when the provider, account, and build support it. It uses your own supported provider connection and sends the sample audio to that provider. Review the sample and consent prompt before creating a voice. See custom voices.

If a preview does not play, check output volume and selected speaker, then try a built-in voice. For microphone or recognition problems, see voice troubleshooting.