Skip to main content

Voice Input

Sometimes typing isn't convenient—you're cooking, exercising, or just want to think out loud. Osaurus includes fully local speech-to-text powered by WhisperKit. Speak naturally, see your words appear in real-time, and know that nothing leaves your Mac.

Features​

  • Real-time transcription — See your words as you speak
  • Multiple Whisper models — From Tiny (75 MB) to Large V3 (3 GB)
  • Microphone or system audio — Transcribe your voice or computer audio
  • Configurable sensitivity — Adjust for quiet or noisy environments
  • Auto-send with confirmation — Hands-free message sending
  • Transcription Mode — Dictate into any app with a global hotkey

Setup​

1. Open Voice Settings​

  1. Open the Management window (⌘⇧M)
  2. Navigate to the Voice tab

2. Grant Microphone Permission​

When prompted, allow Osaurus to access your microphone:

  1. Click Enable Microphone in Voice settings
  2. macOS will show a permission dialog
  3. Click Allow

If you accidentally denied permission:

  1. Open System Settings → Privacy & Security → Microphone
  2. Find Osaurus in the list
  3. Toggle the switch to enable

3. Download a Whisper Model​

Choose and download a transcription model:

ModelSizeSpeedAccuracyBest For
Tiny75 MBFastestBasicQuick notes, casual use
Base142 MBFastGoodGeneral purpose
Small466 MBModerateBetterBalanced performance
Medium1.5 GBSlowerVery GoodHigher accuracy needs
Large V33 GBSlowestBestMaximum accuracy, complex audio

Click Download next to your chosen model. The model downloads to ~/.osaurus/whisper-models.

Recommended

Start with Small for a good balance of speed and accuracy. Upgrade to Large V3 if you need better transcription for accented speech or technical vocabulary.

4. Test Your Voice Input​

  1. Click the microphone button in the chat input area
  2. Speak a test phrase
  3. Watch the transcription appear in real-time
  4. Click the microphone button again to stop

Using Voice Input​

Starting Voice Input​

  • Click the microphone icon in the chat input area
  • The icon animates to show recording is active
  • Speak naturally at a normal pace

Stopping Voice Input​

  • Click the microphone icon again
  • Or wait for the silence timeout
  • Transcription is inserted into the input field

Auto-Send​

When enabled, voice input automatically sends your message after transcription:

  1. Open Voice settings (⌘⇧M → Voice)
  2. Enable Auto-send after transcription
  3. Optionally enable Confirm before sending for a review step

Switching Audio Sources​

Choose between microphone and system audio:

SourceDescriptionUse Case
MicrophoneYour voice inputDictating messages
System AudioAudio playing on your computerTranscribing videos, meetings

Select the audio source in Voice settings.

Sensitivity Settings​

Adjust voice detection for your environment:

SettingDescription
SensitivityHow loud audio must be to trigger transcription
Silence TimeoutHow long to wait after silence before stopping
  • Quiet environment — Lower sensitivity, shorter timeout
  • Noisy environment — Higher sensitivity, longer timeout

VAD Mode (Voice Activity Detection)​

VAD Mode enables hands-free activation of Osaurus using wake phrases. Say your persona's name or a custom phrase to open chat and start voice input automatically.

Enabling VAD Mode​

  1. Open Voice settings (⌘⇧M → Voice)
  2. Toggle VAD Mode to enable
  3. Configure wake phrases (optional)

How VAD Works​

  1. Always Listening — Osaurus listens for wake phrases in the background
  2. Activation — When a wake phrase is detected, chat opens automatically
  3. Voice Input Starts — Microphone activates for your message
  4. Automatic Close — Chat closes after silence timeout (configurable)

Status Indicators​

IndicatorMeaning
Blue pulsing dotVAD is active and listening for wake phrases
Microphone icon activeCurrently recording voice input

The blue pulsing dot appears on the menu bar icon when VAD is enabled and listening.

Wake Phrases​

By default, saying a persona's name activates that persona:

  • "Hey Code Assistant" → Opens chat with Code Assistant persona
  • "Hey Research Helper" → Opens chat with Research Helper persona

You can also configure custom wake phrases in Voice settings.

VAD Settings​

SettingDescription
Wake PhrasesCustom phrases to trigger activation
Silence TimeoutTime to wait after silence before closing
Auto-CloseAutomatically close chat after inactivity

VAD Use Cases​

  • Hands-free computing — Control Osaurus while cooking, exercising, etc.
  • Accessibility — Voice-first interaction
  • Quick queries — Fast questions without touching the keyboard
  • Multi-tasking — Keep working while asking questions

Transcription Mode​

Transcription Mode lets you dictate text directly into any application on your Mac. Press a global hotkey, speak, and watch your words appear in the currently focused text field—no copy-paste required.

What is Transcription Mode?​

Unlike regular voice input (which types into Osaurus chat), Transcription Mode types into whatever app you're using. Writing an email? Dictate it. Filling out a form? Speak instead of type. Coding? Dictate comments without switching windows.

Features​

  • Global Hotkey — Trigger transcription from anywhere on your Mac
  • Live Typing — Text appears in the focused text field in real-time
  • Accessibility Integration — Uses macOS accessibility APIs to simulate keyboard input
  • Minimal Overlay — Sleek floating UI shows recording status
  • Press Esc or Done — Stop transcription when finished

Setup​

1. Grant Accessibility Permission​

Transcription Mode requires accessibility access to type into other applications:

  1. Open System Settings → Privacy & Security → Accessibility
  2. Click the + button
  3. Find and add Osaurus
  4. Toggle the switch to enable
Why Accessibility Access?

Transcription Mode simulates keyboard input to type into other apps. macOS requires explicit permission for any app that controls input in other applications.

2. Configure the Global Hotkey​

  1. Open Management window (⌘⇧M) → Voice
  2. Navigate to the Transcription tab
  3. Click the hotkey recorder field
  4. Press your desired key combination (e.g., ⌥Space)
  5. The hotkey is saved automatically

Using Transcription Mode​

  1. Focus a text field — Click into any text input (email, document, search bar, code editor)
  2. Press your hotkey — The transcription overlay appears
  3. Speak — Your words are typed into the focused field in real-time
  4. Stop — Press Esc or click Done to finish

The Transcription Overlay​

When active, a minimal overlay shows:

ElementDescription
Recording indicatorVisual confirmation that you're being transcribed
Done buttonClick to stop transcription
Audio levelReal-time visualization of your voice

The overlay stays out of your way while giving you clear feedback.

Tips for Best Results​

  1. Speak at a natural pace — No need to slow down or over-enunciate
  2. Pause for punctuation — Brief pauses help the model place periods and commas
  3. Use a quiet environment — Background noise reduces accuracy
  4. External mic recommended — Better audio quality improves transcription

Transcription Mode Use Cases​

  • Email and messaging — Dictate messages in any app
  • Document writing — Draft content in Word, Pages, Google Docs, etc.
  • Code comments — Speak documentation without switching to Osaurus
  • Form filling — Voice input for any text field on the web
  • Note-taking — Quick capture in any notes app
  • Search — Speak search queries in Spotlight, browsers, or any app

Troubleshooting​

Microphone Not Working​

  1. Check microphone permission in System Settings → Privacy & Security → Microphone
  2. Verify the correct audio input is selected in Voice settings
  3. Test your microphone in another app
  4. Restart Osaurus

Poor Transcription Quality​

  1. Upgrade your model — Try a larger Whisper model
  2. Reduce background noise — Move to a quieter environment
  3. Speak clearly — Maintain consistent volume and pace
  4. Check microphone quality — External mics often work better

VAD Not Activating​

  1. Verify VAD is enabled in Voice settings
  2. Check the blue pulsing dot is visible on the menu bar icon
  3. Speak the wake phrase clearly
  4. Adjust sensitivity settings
  5. Ensure microphone permissions are granted

High CPU Usage​

Whisper transcription is computationally intensive:

  1. Use a smaller model — Tiny or Base for lower resource usage
  2. Close unnecessary apps — Free up system resources
  3. Disable VAD when not needed — Always-listening uses continuous CPU

Model Download Failed​

  1. Check your internet connection
  2. Verify disk space (Large V3 needs 3 GB)
  3. Try a smaller model first
  4. Check ~/.osaurus/whisper-models for partial downloads and delete them

Transcription Mode Not Working​

  1. Check accessibility permission — Open System Settings → Privacy & Security → Accessibility and verify Osaurus is enabled
  2. Verify the hotkey — Ensure your global hotkey doesn't conflict with other apps
  3. Focus a text field — Transcription Mode requires an active text input
  4. Restart Osaurus — Some permission changes require a restart

Text Not Appearing in Target App​

  1. Ensure the text field is focused — Click into the field before starting
  2. Check app compatibility — Some apps with custom text fields may not work
  3. Try a different app — Test with a standard app like TextEdit to verify functionality
  4. Grant accessibility again — Remove and re-add Osaurus in accessibility settings

Privacy​

Voice input in Osaurus is completely local:

  • No cloud processing — All transcription happens on your Mac
  • No data uploaded — Audio never leaves your device
  • Models stored locally — Downloaded once, used offline
  • VAD is local — Wake phrase detection runs on-device

For persona-specific voice workflows, see the Personas guide.