User guide

OneSentence lives in the menu bar. Default hotkeys work in any app. Current release notes are on the changelog.

Voice to text (Listen)

1. Press ⌘5 to start recording. The menu bar icon shows a recording badge.
2. Speak into the selected microphone. Parakeet is the default engine (European languages). Whisper is optional for broader coverage. Default recording limit is 300 seconds (change this in Settings).
3. Press ⌘5 again to stop. Processing runs on-device.
4. The transcript is copied to the clipboard and, by default, pasted into the field you were typing in. Turn that off in Settings → General if you only want the clipboard. Insert needs Accessibility permission.

Text to speech (Read)

Supertonic voices are English-only.

1. Select text in any application.
2. Press ⌘6. OneSentence copies the selection, speaks it, and restores the previous clipboard.
3. Other apps can duck in volume while speech is playing. The session can appear in macOS Now Playing.
4. Press ⌘6 again to stop early.

Snippets (Universal Palette)

1. Open Settings → Snippets and choose a folder. OneSentence watches that folder.
2. Put plain files there: .txt, .md, .snippet, or Emacs .yasnippet.
3. Press ⌘⇧Space anywhere. Type to fuzzy-search, then Enter to paste. Clipboard history is preserved.

Local history

If History is enabled, each transcription is stored in a local database on your Mac. Open History from the menu bar, search, and click an entry to copy it. You can delete individual rows or clear the database. Nothing is synced to a cloud account.

Useful options

• Refine punctuation: normalize periods, quotes, and dashes before speech.
• Recording timeout: 10 seconds to 10 minutes (default 300 seconds).
• Read transcription back: speak the transcript immediately after Listen finishes.
• Insert into the active field: paste the transcript where you were typing (on by default).

Settings reference

General

• Launch at Login: keep the menu bar extra available after reboot.
• Shortcuts: change Listen (default ⌘5) and Read Selection (default ⌘6).
• Speaking rate: 0.5x to 2.0x (synthesis tempo).
• Voice quality: Fast, Balanced, or Quality. More steps than Quality do not help this voice.
• Repair encoding errors: fix garbled characters after transcription.
• Sound effects and visual alerts: confirmation tones and a status HUD.
• Insert transcription into the active field: paste after Listen (on by default; needs Accessibility).

Voices

Ten English Supertonic voices (five male, five female) are available after the model download. Switching voices does not require a further download.

Models

• Engine: Parakeet (default, faster TDT) or Whisper (optional, more languages). Whisper does not download until you tap Download on a model.
• Whisper language: Auto-Detect by default. Pin a language only if auto is wrong.
• Whisper acceleration on macOS 14+: Apple Neural Engine or Metal. ANE uses a separate model package.
• Language: Auto, or pin a language for speed and fewer mis-detects.
• On-demand download: only the models you select are fetched and verified.
• Background prewarming: initialize the engine at launch to shorten the first capture.

Audio

Choose input and output devices without restarting. Optional ducking lowers other apps while OneSentence is speaking or recording.

Maintenance

• Reset Settings: restore preferences and hotkeys; keep downloaded models.
• Factory Reset: delete downloaded models, settings, and hotkeys. The app bundle stays.
• Debug logging: write logs and recorded audio to the application data folder for troubleshooting.

Source and licenses

Source and releases: github.com/snowy-0wl/onesentence-app. Third-party model licenses: CREDITS.md. Site license page: License.