Troubleshoot voice

Voice input and voice output use separate paths. First check whether Nock reacted to the shortcut, showed a transcript, produced an answer, and played speech. Testing each step separately avoids repeating a long request to diagnose a missing voice.

Test the parts separately

1. Type a short question in Nock. This checks whether the agent can answer.
2. Make a short dictation and look for a caption. This checks microphone capture and speech recognition.
3. Open Settings → Agent → Agent voice and preview a voice. This checks speech playback.

The shortcut does not react

Check the current binding under Settings → General → Shortcuts and try the keys shown there. On Mac, check keyboard permissions in Settings → Permissions. On Windows, use the platform shortcut shown in settings. If the keys start a recording but no words appear, continue with the microphone checks below. See keyboard shortcuts.

Recording starts, but no words appear

Confirm the expected microphone is selected in the operating system and that its input meter moves while you speak. Check Nock’s microphone access in Settings → Permissions. On Mac, Nock Speech may have its own microphone entry. Then check the status under Settings → Dictation → Where your voice is turned into text. The selected local model may need downloading or may not support the chosen language; cloud recognition needs an available account route or supported own key.

On Mac, Apple’s recognizer also needs its speech-recognition access and on-device language support. Nock’s local Nock model is unavailable on macOS 13. See local dictation.

An answer appears but is not spoken

Check Enable Agent voice, the selected route, and the computer’s output volume/device. Try a preview. If the voice is local, read the model status: a missing or unsupported model can result in a fallback voice or silent playback. If cloud voice is selected, check the account/provider message and network access. A preview can work while a separate agent request fails, and vice versa.

Speech stops early or sounds different

Starting a new recording, sending another request, or pressing a stop control can interrupt speech. The spoken line may be shorter than the complete answer shown on screen. If only the local natural voice has trouble keeping up, the app can fall back to a simpler voice or system speech. Read the visible answer when exact wording matters.

Report the failing step

Include app version, operating system, input or output device, the route selected in settings, and what happened in the three tests above. Share only a short non-sensitive example. Do not attach recordings, private conversations, keys, or screenshots unless they are necessary and you have removed unrelated information.