Dictate straight into a note — transcription only, with a small set of spoken commands to control the session, not to format the note.
Start voice input from the command palette (Start voice input) to start listening. Speech accumulates in an editable preview box instead of the document itself, so you can fix a mishear before it ever touches your note. Say "insert this" (or press Enter) to commit the reviewed text at the cursor, Shift+Enter to add a line break within the preview, or Escape to discard it.
Voice input uses the browser's built-in Web Speech API, which only Chromium-based browsers implement — see
Installation. The mic button is hidden entirely on unsupported browsers.
Dictation is transcription only — it never inserts Markdown syntax or otherwise formats the note from speech. A small set of session-control phrases works instead, since those act on the preview buffer rather than the document:
"new paragraph" / "new line" / "new row"
"period" / "comma" / "question mark" / "exclamation point"
"colon" / "semicolon"
"scratch that" / "delete last" / "undo that"
"scratch all text" / "clear all text" / "clear everything"
"insert this/it/text" / "commit this/it/text"
"insert this and stop listening"
"stop listening" / "done listening"
Everything else is transcribed as plain text, so ordinary dictation always works. Sentences capitalize themselves automatically after a spoken "period", "question mark", or "exclamation point" (a comma or colon doesn't count).
Listening stops automatically when the pane loses focus, the tab is backgrounded, or the mic button is clicked again. If the browser denies microphone access, loses its network connection mid-session, or can't find a microphone, a toast explains why and listening stops rather than retrying silently.