Voice Input

Dictate messages and text using push-to-talk voice recognition.

Desktop only. Voice input is not available in the web app.

Voice input lets you dictate text instead of typing. Use it in the AI chat input, the Markdown editor, or the floating assistant to compose messages and content hands-free.

How to Use Voice Input

Voice input is push-to-talk — hold the shortcut to record, release to stop and transcribe.

  1. Focus the input field where you want to dictate (chat input, editor, or floating assistant)
  2. Hold Cmd+R (Windows: Ctrl+R)
  3. Speak your message
  4. Release the keys — Omnilib transcribes and inserts the text at the cursor

The input field shows a pulsing indicator while recording is active.

Platform Recognition Engines

Omnilib uses the operating system's native speech recognition for low-latency streaming transcription:

macOS — Uses Apple's SFSpeechRecognizer framework. Transcription streams in real time as you speak, with confidence scoring. Omnilib shows lower-confidence words with a subtle highlight so you can review them quickly after dictating.

Windows — Uses the Windows Runtime Windows.Media.SpeechRecognition API. Transcription is delivered at the end of each utterance.

Both engines run on-device. Audio is not sent to external servers.

Language Detection

Voice input automatically uses the language you have selected in the app's language switcher. Supported languages:

  • English
  • Italian
  • Spanish

Switch the app language in the top toolbar to change the recognition language.

Where Voice Input Works

You can use voice input in:

  • AI chat input — Dictate messages before sending
  • Markdown editor — Dictate paragraphs, headings, and content directly into the document
  • Floating assistant — Dictate into the compact chat input when the floating panel is open

Voice input inserts text at the current cursor position. If no text is selected, it appends to the end of any existing content in single-line inputs.

Tips for Accurate Transcription

  • Speak at a natural pace — pausing slightly between sentences helps the recognizer segment speech correctly
  • State punctuation explicitly if needed: "comma", "period", "new paragraph"
  • If a word is transcribed incorrectly, click it in the input field and type the correction — voice input does not replace existing text after transcription is complete
  • In noisy environments, use a headset or directional microphone for better accuracy

Related