Skip to main content

​
General Questions

VoiceTypr is an open source, AI-powered voice-to-text dictation tool for macOS and Windows. It’s an alternative to proprietary tools like Wispr Flow and SuperWhisper, offering:
  • 100% offline transcription - your voice never leaves your device
  • System-wide hotkey - record anywhere with a single keystroke
  • Automatic text insertion - transcribed text appears at your cursor
  • Multiple AI models - choose between speed and accuracy
  • 99+ languages supported - automatic language detection
  • Open source - full transparency under AGPL-3.0 license
Pay once, use forever - no subscriptions or cloud dependencies.
Yes! VoiceTypr is 100% offline for core transcription:
  • AI models run entirely on your device (CPU or GPU)
  • Your voice recordings never leave your computer
  • No internet connection required for transcription
The only internet connections are for:
  • Initial model downloads (~75 MB to 2.9 GB, one-time per model)
  • Update checks (optional, checks GitHub for new versions)
  • License validation (trial check only, with 90-day offline grace period)
  • AI Enhancement (optional feature using Groq/Gemini/OpenAI APIs)
All other operations work completely offline, including recording, transcription, and text insertion.
Absolutely. VoiceTypr is privacy-first by design:
  • No cloud uploads - all transcription happens on your device
  • No telemetry or tracking - we don’t collect usage data
  • No third-party analytics - zero external services
  • Open source - you can verify the code yourself
Your recordings are:
  • Stored temporarily during transcription
  • Automatically deleted after processing
  • Never sent to any server
The only data that leaves your device is:
  • Update checks to GitHub (just version number)
  • Optional AI Enhancement (if you enable it and provide API keys)
FeatureVoiceTyprSuperWhisperWispr Flow
PricePay onceSubscriptionSubscription
Open Source✅ Yes (AGPL-3.0)❌ No❌ No
Offline✅ 100%✅ Yes✅ Yes
PlatformsmacOS, WindowsmacOS onlymacOS, Windows
GPU AccelerationMetal, Vulkan, ANEMetal onlyMetal only
AI ModelsWhisper, ParakeetWhisperWhisper
AI EnhancementGroq, Gemini, OpenAIBuilt-inBuilt-in
Custom Hotkeys✅ Yes✅ Yes✅ Yes
Multi-language99+ languages99+ languages99+ languages
Key advantages of VoiceTypr:
  • Open source - full transparency and community contributions
  • One-time payment - no recurring costs
  • Windows support with Vulkan GPU acceleration
  • Apple Neural Engine support via Parakeet models
  • Community-driven development

​
Pricing & Licensing

VoiceTypr uses a pay-once, use forever licensing model:
  • Free trial: Full-featured trial period
  • Lifetime license: One-time payment for permanent access
  • No subscriptions: Pay once, own it forever
  • Free updates: All future updates included
Important: VoiceTypr is open source (AGPL-3.0), so you can also:
  • Build from source for free (for personal use)
  • Contribute to development
  • Fork and modify (must remain open source)
Commercial use of pre-built binaries requires a license.Check the website for current pricing.
VoiceTypr has a 90-day offline grace period after trial ends:
  • You can continue using VoiceTypr offline for 90 days
  • Internet connection required to activate a license
  • Trial status is checked on launch (requires internet)
  • After grace period, you must activate or VoiceTypr will be limited
To activate after trial:
  1. Purchase a license key
  2. Open VoiceTypr → Settings → License
  3. Enter your license key
  4. Click “Activate”
Your license is tied to your device ID and works offline indefinitely after activation.
License terms vary by purchase option. Typically:
  • Single license: One device
  • Multi-device licenses: Available for purchase
  • Device ID: Tied to hardware (OS metadata + system info)
To transfer license:
  1. Deactivate on old device (Settings → License → Deactivate)
  2. Activate on new device with same license key
Contact support for multi-device licensing options.
Yes! VoiceTypr is licensed under AGPL-3.0:
  • Source code available: GitHub repository
  • Build from source: Instructions in README
  • Contributions welcome: Pull requests accepted
  • Fork allowed: Must remain open source (AGPL-3.0)
What this means:
  • You can inspect all code
  • You can build for personal use without a license
  • Commercial use of pre-built binaries requires a license
  • Modifications must also be open sourced
This ensures transparency and privacy while supporting sustainable development.

​
Platform Support

macOS:
  • macOS 13.0 (Ventura) or later
  • Intel and Apple Silicon (M1/M2/M3+) Macs
  • DMG installer available
  • Exclusive Parakeet model support (Apple Silicon only)
Windows:
  • Windows 10 (64-bit) or Windows 11
  • x64 and ARM64 architectures
  • NSIS installer with GPU detection
  • Vulkan GPU acceleration support
Linux: Not currently supported (contributions welcome!)iOS/Android: Not planned (requires different architecture)
Yes! VoiceTypr supports both Intel and Apple Silicon Macs:Intel Macs:
  • ✅ Full Whisper model support
  • ✅ Metal GPU acceleration
  • ✅ All core features
  • ❌ No Parakeet models (requires Apple Neural Engine)
Apple Silicon Macs (M1/M2/M3+):
  • ✅ All Whisper models with Metal
  • ✅ Exclusive Parakeet model support (Apple Neural Engine)
  • ✅ Better performance and battery efficiency
Performance: Apple Silicon is 2-5x faster with better battery life, but Intel Macs work great with Whisper models.
Yes! VoiceTypr for Windows supports all Vulkan-compatible GPUs:Supported GPUs:
  • ✅ NVIDIA: GeForce GTX 900+ (Vulkan 1.0+)
  • ✅ AMD: Radeon RX 400+ (Vulkan 1.0+)
  • ✅ Intel: Arc, Iris Xe, UHD Graphics 600+ (Vulkan 1.0+)
Performance boost: 5-10x faster transcription with GPU accelerationRequirements:
  • Up-to-date graphics drivers (includes Vulkan Runtime)
  • 1+ GB VRAM
Automatic fallback: If GPU is unavailable, VoiceTypr uses CPU mode automatically.See Windows Platform Guide for driver update links.

​
Features & Usage

VoiceTypr supports 99+ languages out of the box through Whisper models:
  • Auto-detect: Automatically identifies the spoken language
  • Manual selection: Force a specific language for better accuracy
  • Multilingual: Switch between languages without restarting
Popular languages include: English, Spanish, French, German, Italian, Portuguese, Dutch, Russian, Chinese (Mandarin), Japanese, Korean, Arabic, Hindi, and many more.How it works:
  1. Speak in any supported language
  2. Whisper automatically detects the language
  3. Transcription appears in the original language
Note: Language detection is more accurate with larger models (Small, Medium, Large).
Whisper Models (All platforms):
ModelSizeSpeedAccuracyBest For
Tiny75 MBFastestLowerQuick notes, fast systems
Base142 MBFastGoodBalanced performance
Small466 MBModerateHighRecommended for most users
Medium1.5 GBSlowVery HighHigh accuracy needs
Large2.9 GBSlowestHighestProfessional use, best quality
Parakeet Models (macOS Apple Silicon only):
ModelSizeSpeedAccuracyEngine
Parakeet TDT 0.6B v3500 MBReal-timeHighApple Neural Engine
Choosing a model:
  • Speed: Use Tiny or Base
  • Accuracy: Use Medium or Large
  • Balance: Use Small (recommended)
  • Battery efficiency (Apple Silicon): Use Parakeet
You can download multiple models and switch between them.
Push-to-Talk (PTT):
  • Hold your hotkey to record
  • Release hotkey to stop and transcribe
  • Like a walkie-talkie
  • Minimum duration: 1 second
  • Best for: Quick, controlled recordings
Toggle Mode:
  • Press hotkey once to start recording
  • Press again to stop and transcribe
  • Or press Esc twice to cancel
  • Minimum duration: 3 seconds
  • Best for: Longer recordings, hands-free
Switch modes:
  • Settings → General → Recording Mode
  • Or menubar/tray → Recording Mode
Tips:
  • PTT is more precise for short bursts
  • Toggle is better for lengthy dictation
  • Both modes work in any application
Yes! VoiceTypr supports audio file upload and transcription:Supported formats:
  • WAV, MP3, M4A, FLAC, OGG, and more
  • Multi-channel audio (up to 4+ channels)
  • Any sample rate (automatically normalized to 16kHz)
How to use:
  1. Open VoiceTypr Settings
  2. Navigate to Upload section
  3. Drag & drop audio file or click to browse
  4. Wait for transcription
  5. Text appears in the upload section (can copy to clipboard)
Note: File transcription does not auto-insert text at cursor - you must copy it manually.Use cases:
  • Transcribe meeting recordings
  • Convert voice memos to text
  • Transcribe podcast segments
  • Process existing audio archives
AI Enhancement is an optional feature that improves transcriptions using cloud AI APIs:What it does:
  • Fixes grammar and punctuation errors
  • Removes filler words (um, uh, like)
  • Formats text appropriately
  • Applies custom prompts (email, commit messages, notes)
Supported providers:
  • Groq (fast, affordable)
  • Gemini (Google’s AI)
  • OpenAI (GPT models)
  • OpenAI-compatible endpoints (custom URLs)
Requirements:
  • Internet connection
  • API key from provider (you provide your own)
  • Costs based on provider’s pricing
How to enable:
  1. Settings → AI Enhancement
  2. Choose provider
  3. Enter API key
  4. Select enhancement mode (or create custom prompt)
  5. Enable enhancement toggle
Privacy: Your transcriptions are sent to the AI provider for enhancement. If privacy is a concern, keep AI Enhancement disabled.

​
Performance & Requirements

Minimum: 3-4 GB free spaceBreakdown:
  • VoiceTypr app: ~200 MB
  • Single Whisper model: 75 MB to 2.9 GB
  • Parakeet model (macOS): ~500 MB
  • Temporary recording files: ~100 MB
Recommended: 5+ GB to have room for multiple modelsStorage locations:
  • macOS: ~/Library/Application Support/com.voicetypr.app/
  • Windows: %APPDATA%\com.voicetypr.app\
Free up space:
  • Delete unused models (Settings → Models → Remove)
  • Use smaller models (Tiny/Base instead of Large)
  • Reset App Data to clear cache (Settings → Advanced)
No, but it helps significantly.Without GPU (CPU only):
  • ✅ VoiceTypr works fine
  • ⚠️ Transcription is 1-2x real-time (depends on CPU)
  • ⚠️ Slower with larger models
  • ✅ No special drivers needed
With GPU:
  • ✅ 5-10x faster transcription (Windows Vulkan)
  • ✅ 2-5x faster (macOS Metal)
  • ✅ Near real-time with Parakeet (Apple Neural Engine)
  • ⚠️ Requires updated graphics drivers
GPU requirements:
  • macOS: Any Mac with Metal support (all modern Macs)
  • Windows: NVIDIA, AMD, or Intel GPU with Vulkan support
Bottom line: VoiceTypr always works - GPU just makes it faster.
Speed depends on:
  1. Platform (macOS vs Windows)
  2. Hardware (GPU vs CPU, Apple Silicon vs Intel)
  3. Model size (Tiny vs Large)
Approximate speeds:Apple Silicon + Parakeet: ~1.0x real-time (30s audio = 30s to transcribe)macOS Metal GPU:
  • Tiny/Base: 5-10x real-time (30s audio = 3-6s)
  • Small/Medium: 2-5x real-time (30s audio = 6-15s)
  • Large: 1-2x real-time (30s audio = 15-30s)
Windows Vulkan GPU:
  • Tiny/Base: 8-12x real-time (30s audio = 2-4s)
  • Small/Medium: 4-8x real-time (30s audio = 4-8s)
  • Large: 2-4x real-time (30s audio = 8-15s)
CPU only (no GPU):
  • Tiny/Base: 1-3x real-time
  • Small/Medium: 0.5-1.5x real-time (slower than real-time)
  • Large: 0.2-0.5x real-time (very slow)
Recommendation: For best experience, use Small model with GPU acceleration.
Battery impact varies by platform and usage:macOS Apple Silicon + Parakeet:
  • ⚡ Minimal impact (Apple Neural Engine is ultra-efficient)
  • Comparable to normal typing
  • Best option for laptops
macOS Intel + Whisper Metal:
  • ⚠️ Moderate impact during active transcription
  • GPU and CPU both active
  • Drains faster than Parakeet
Windows GPU (Vulkan):
  • ⚠️ Moderate to high impact during transcription
  • GPU acceleration uses more power than CPU alone
  • Worth it for speed boost
CPU-only mode:
  • ⚠️ High impact due to sustained CPU usage
  • Slower transcription = longer CPU time = more drain
Battery saving tips:
  1. Use Parakeet on Apple Silicon Macs
  2. Use smaller models (Tiny/Base)
  3. Keep recordings short
  4. Disable AI Enhancement (requires internet/API calls)
  5. Quit VoiceTypr when not in use

​
Troubleshooting

See the Troubleshooting Guide for detailed solutions on slow transcription.Quick fixes:
  1. Update graphics drivers (Windows - most common fix!)
  2. Use smaller models (Tiny/Base instead of Large)
  3. Try Parakeet (macOS Apple Silicon only)
  4. Check GPU detection in logs
  5. Free up RAM by closing background apps
See the Troubleshooting Guide for detailed solutions on recording issues.Quick checklist:
  1. ✅ Microphone permission granted
  2. ✅ At least one model downloaded
  3. ✅ Hotkey not conflicting with other apps
  4. ✅ VoiceTypr running in background (menubar/tray icon visible)
See the Troubleshooting Guide for detailed solutions on improving transcription accuracy.Quick improvements:
  1. Use larger models (Small, Medium, or Large)
  2. Reduce background noise
  3. Speak clearly at normal pace
  4. Set language explicitly instead of auto-detect
  5. Enable AI Enhancement for grammar fixes (requires API key)
Documentation:Community Support:Before asking for help:
  1. Check Troubleshooting first
  2. Search existing GitHub issues
  3. Collect diagnostic info (Settings → Help → Copy System Info)
  4. Include logs if relevant (Settings → Help → View Logs)
Response time: Community-driven project - response times vary. Be patient and provide details!

​
See Also