Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.voicetypr.com/llms.txt

Use this file to discover all available pages before exploring further.

​
Overview

VoiceTypr is fully optimized for macOS with native features including Metal acceleration, Apple Neural Engine support, and seamless system integration.

​
System Requirements

​
Minimum Requirements

  • macOS Version: 13.0 (Ventura) or later
  • Free Disk Space: 3-4 GB (for AI models)
  • Memory: 8 GB RAM recommended
  • Processor: Apple Silicon or Intel Mac
  • macOS Version: 14.0 (Sonoma) or later
  • Processor: Apple Silicon (M1, M2, M3, or newer) for best performance
  • Free Disk Space: 5+ GB for multiple models

​
Installation

​
From DMG

  1. Download the latest VoiceTypr.dmg from GitHub Releases
  2. Open the DMG file
  3. Drag VoiceTypr to your Applications folder
  4. Launch VoiceTypr from Applications
  5. Follow the onboarding wizard to:
    • Grant required permissions
    • Download your preferred AI model
    • Set up your recording hotkey
VoiceTypr is fully signed and notarized by Apple, so you can run it without security warnings.

​
macOS-Specific Features

​
NSPanel Integration

VoiceTypr uses NSPanel for the recording pill window, which:
  • Floats above all windows without stealing focus
  • Appears when recording starts
  • Provides real-time visual feedback
  • Doesn’t disrupt your workflow

​
Metal Acceleration

When using Whisper models, VoiceTypr leverages Metal for GPU-accelerated transcription:
  • Automatically uses Metal on all Macs (Intel and Apple Silicon)
  • Falls back to CPU if Metal initialization fails
  • 2-5x faster transcription than CPU-only mode
  • Zero configuration required

​
Apple Neural Engine (Parakeet Models)

For Apple Silicon Macs, VoiceTypr offers exclusive access to Parakeet models:
  • Native CoreML implementation via FluidAudio SDK
  • Runs on the Apple Neural Engine for exceptional efficiency
  • Lightweight 1.2MB Swift sidecar (vs 123MB Python alternative)
  • On-demand model download (~500MB)
  • Optimized for real-time transcription
  • macOS 13.0+ required
Parakeet models are only available on macOS with Apple Silicon. Intel Macs and other platforms use Whisper models.

​
Required Permissions

VoiceTypr requires two system permissions on macOS:

​
1. Microphone Access

Required for: Recording your voice
  • Prompt appears on first recording attempt
  • Navigate to: System Settings → Privacy & Security → Microphone
  • Enable for VoiceTypr

​
2. Accessibility Access

Required for: Automatic text insertion at cursor position
  • Navigate to: System Settings → Privacy & Security → Accessibility
  • Click the + button
  • Add VoiceTypr from Applications
  • Toggle the checkbox to enable
Without Accessibility permission, VoiceTypr cannot automatically insert transcribed text. You’ll need to paste manually.

​
Platform Integration

​
Dock Icon Management

VoiceTypr intelligently manages its Dock icon:
  • Hidden when only the menubar is visible
  • Shown when the main window is open
  • Reduces clutter while maintaining accessibility

​
Media Remote Integration

On macOS, VoiceTypr integrates with system media controls:
  • Recording status visible in media remote
  • Provides context for active recording sessions
The menubar icon provides quick access to:
  • Start/stop recording
  • Recent transcriptions (copy to clipboard)
  • Model selection
  • Recording mode (Push-to-Talk / Toggle)
  • Check for updates
  • Open settings

​
Code Signing & Notarization

All VoiceTypr releases for macOS are:
  • Code signed with a valid Apple Developer ID
  • Notarized by Apple for added security
  • Gatekeeper approved - no security warnings on launch
This ensures a safe, trusted installation experience.

​
Model Management

​
Whisper Models (All Macs)

Supported Whisper models with Metal acceleration:
  • Tiny - Fastest, lower accuracy (~75 MB)
  • Base - Balanced speed/accuracy (~142 MB)
  • Small - Recommended for most users (~466 MB)
  • Medium - High accuracy (~1.5 GB)
  • Large - Best accuracy, slower (~2.9 GB)

​
Parakeet Models (Apple Silicon Only)

  • Parakeet TDT 0.6B v3 - CoreML optimized (~500 MB)
  • Uses Apple Neural Engine for efficient transcription
  • Download on-demand via Settings
  • Cached in ~/Library/Application Support/FluidAudio/
For Apple Silicon Macs, try Parakeet models for the best balance of speed and accuracy with minimal battery impact.

​
Performance Characteristics

​
Apple Silicon (M1/M2/M3+)

  • Whisper (Metal): 2-5x real-time speed
  • Parakeet (ANE): Near real-time with exceptional efficiency
  • Battery Impact: Minimal with Parakeet, moderate with Whisper
  • Memory Usage: 100-500 MB depending on model

​
Intel Macs

  • Whisper (Metal): 1-3x real-time speed
  • Battery Impact: Moderate to high during transcription
  • Memory Usage: 200-600 MB depending on model

​
Troubleshooting

​
App Won’t Open

Issue: “VoiceTypr is damaged and can’t be opened” Solution: This usually indicates Gatekeeper issues. Run: 
xattr -cr /Applications/VoiceTypr.app

​
Metal Acceleration Not Working

Issue: Slow transcription performance Solution:
  1. Check Console.app for Metal errors
  2. Update macOS to the latest version
  3. VoiceTypr automatically falls back to CPU if Metal fails

​
Parakeet Model Download Fails

Issue: Model download stalls or fails Solution:
  1. Ensure you have 2+ GB free disk space
  2. Check your internet connection
  3. Try deleting and re-downloading:
    • Settings → Advanced → Reset App Data
    • Restart VoiceTypr
    • Re-download the model

​
Accessibility Permission Denied

Issue: Text insertion doesn’t work Solution:
  1. Open System Settings → Privacy & Security → Accessibility
  2. Find VoiceTypr in the list
  3. Toggle it off then on again
  4. Restart VoiceTypr

​
Updates

VoiceTypr includes automatic update checking:
  • Checks for updates on launch (requires internet)
  • Downloads updates in the background
  • Notifies when ready to install
  • Menubar → Check for Updates for manual check
Update checks are the only internet connection required for normal operation. All transcription happens offline.

​
Uninstalling

To completely remove VoiceTypr:
  1. Quit VoiceTypr from the menubar
  2. Delete the app: Drag VoiceTypr from Applications to Trash
  3. Remove app data (optional): 
    rm -rf ~/Library/Application\ Support/com.voicetypr.app
rm -rf ~/Library/Application\ Support/FluidAudio
rm -rf ~/Library/Caches/FluidAudio
rm -rf ~/Library/Preferences/com.voicetypr.app.plist

​
See Also