Dictation

Speak into the focused app, choose where recognition runs, and recover text when insertion misses.

Hold your dictation shortcut, speak, and release. Nock recognizes the recording, applies your dictation cleanup and replacements, then pastes the finished text into the app in front. It does not type each word into the target app as you speak.

A first dictation

Click inside a real text field in the app where you want the words.
Hold the Dictation Mode shortcut shown in Settings → General. The default is Fn / Globe on Mac and Ctrl + Win on Windows.
Speak while the live caption appears in Nock.
Release the shortcut and wait for recognition and paste to finish.
Review the inserted words before sending the message, running a command, or submitting a form.

Nock listens briefly after release to catch the end of a word. Press Escape to cancel. A very short touch does not start a recording. For hands-free dictation, double-tap your dictation shortcut or use its hands-free shortcut; see keyboard shortcuts.

In Settings → Dictation → Advanced, Stop listening when you go quiet is on by default. After you release the keys, Nock can stop once your voice has ended or the recognizer has returned its final words, up to one second later. Turn it off if the end of a softly spoken word is being cut off; Nock then listens for the full second after release.

Choose where speech is recognized

Open Settings → Dictation → Where your voice is turned into text:

Choice	What happens
Automatic	Local first. Cloud may be used when needed and available for your account.
On this computer	Uses an installed local recognizer. Audio stays on the computer; the recognizer may need a model download.
Cloud	Sends audio to the configured speech route for recognition. You can use an eligible account route or your own supported speech key.

Turn on Never send my voice off this computer to make dictation refuse instead of sending audio to cloud speech if local recognition cannot run. See local dictation for model and platform details. Automatic recognition does not switch routes in the middle of a recording.

Review and recover a paste

Nock pastes into the app that is in front when dictation finishes. It cannot confirm that the intended text field accepted the paste. If focus changes, pause before typing more. When Nock shows a clipboard notice, click the correct field and paste with the platform’s normal paste shortcut. You can also use Paste Last Transcript before starting another recording; it keeps only the latest finished recording in memory until Nock quits.

Dictation can paste into a terminal just like any other text field. Review terminal text carefully: Nock does not decide whether dictated text is safe to execute.

Clean up the result

Settings → Dictation → Polishing offers Off, Light, and (when available) Polished. Light uses local rules. Polished sends the recognized text for a short online cleanup request and falls back to Light if that request cannot be used. Your Your words list can help with names, and When I say … Type … replacements expand exact phrases. To speak into Nock’s own chat field or widget builder, use that composer’s mic button. Show mic button in Nock’s composers is on by default in Settings → Dictation → Advanced; turn it off to hide those buttons. Recognized words remain a draft until you choose to send. Read dictation cleanup before using Polished with sensitive text.

Limits and troubleshooting

The desktop hold has a 20-minute ceiling and warns shortly before it stops. This ceiling is not a promise that a hosted speech service accepts a recording that long. Shorter recordings are easier to review and troubleshoot.

If nothing reacts, check the shortcut and permissions. If the caption is missing, check local dictation, your microphone input, and voice troubleshooting. If the caption is right but the target field is empty, use dictation troubleshooting.