Compatibility
Platform Compatibility
Supported platforms, requirements, and feature compatibility for macOS, Windows, and Linux.
Supported Platforms
| Platform | Status | Notes |
|---|---|---|
| macOS Apple Silicon (M1, M2, M3, M4) | Available | Best experience, native GPU acceleration |
| macOS Intel (x86_64) | Available | Fully functional, CPU-based transcription |
| Windows 10+ (x64) | In Development | Core components compatible, adaptation in progress |
| Linux | Not Supported | No immediate plans |
macOS Apple Silicon (M1, M2, M3, M4)
Overview
Apple Silicon Macs provide the optimal Vocoding experience. The native GPU acceleration makes transcription with large Whisper models fast and efficient.
Requirements
| Requirement | Minimum | Recommended |
|---|---|---|
| macOS version | 10.15 (Catalina) | 13+ (Ventura) |
| RAM | 4 GB | 8 GB+ |
| Disk | 500 MB + models | 2 GB+ |
| Permissions | Microphone | Microphone + Accessibility |
Feature Compatibility
| Feature | Status | Details |
|---|---|---|
| Audio Recording | Full | Built-in and external microphones |
| Whisper Transcription | Optimal | GPU-accelerated (Metal), all models supported |
| Ollama (Local + Cloud) | Full | Native Apple Silicon builds. Local mode (free, offline) and cloud mode (Free $0, Pro $20/month, Max $100/month) |
| Cloud LLM (Groq/OpenRouter) | Full | Requires internet + API key |
| Global Shortcuts | Full | Option+Space, Cmd+Shift+R, etc. |
| Overlay Window | Full | Always-on-top without stealing focus |
| Auto Updates (OTA) | Full | Built-in update system |
| Secure Key Storage | Full | macOS Keychain |
Recommended Whisper Models
| Priority | Model | Why |
|---|---|---|
| Best overall | large-v3-turbo | Fast + accurate with Metal GPU |
| Best accuracy | large-v3 | Highest accuracy, reasonable speed on M-series |
| Fast drafts | base | Very fast, good for quick notes |
macOS Intel (x86_64)
Overview
Intel Macs are fully supported. The main difference from Apple Silicon is that Whisper transcription runs on CPU instead of GPU, which makes larger models slower.
Requirements
| Requirement | Minimum | Recommended |
|---|---|---|
| macOS version | 10.15 (Catalina) | 12+ (Monterey) |
| RAM | 4 GB | 8 GB+ |
| Disk | 500 MB + models | 2 GB+ |
| Permissions | Microphone | Microphone + Accessibility |
Feature Compatibility
| Feature | Status | Details |
|---|---|---|
| Audio Recording | Full | Built-in and external microphones |
| Whisper Transcription | Full (CPU) | No GPU acceleration; large models will be slow |
| Ollama (Local + Cloud) | Full | Intel Mac builds. Local mode (free, offline) and cloud mode (no local hardware requirements) |
| Cloud LLM (Groq/OpenRouter) | Full | Requires internet + API key |
| Global Shortcuts | Full | Same shortcuts as Apple Silicon |
| Overlay Window | Full | Always-on-top without stealing focus |
| Auto Updates (OTA) | Full | Built-in update system |
| Secure Key Storage | Full | macOS Keychain |
Recommended Whisper Models
| Priority | Model | Why |
|---|---|---|
| Best balance | small | Good accuracy, reasonable CPU speed |
| Higher accuracy | medium | Better accuracy, slower on CPU |
| Fast drafts | tiny or base | Very fast even on CPU |
Performance Notes
- Models
large-v3andlarge-v3-turbowill work but may be significantly slower without GPU acceleration - For real-time or near-real-time use, stick with
smallor lighter models - Transcription speed depends on your specific Intel processor generation
Windows 10+ (x64)
Overview
The Windows version is in active development. Many core components are technically compatible with Windows, but integration and testing are ongoing.
Requirements
| Requirement | Minimum | Recommended |
|---|---|---|
| Windows version | Windows 10 (x64) | Windows 11 |
| Processor | x64 with AVX2 | Modern x64 processor |
| RAM | 4 GB | 8 GB+ |
| Disk | 500 MB + models | 2 GB+ |
| Runtime | WebView2 | Pre-installed on Windows 11 |
Feature Compatibility
| Feature | Status | Details |
|---|---|---|
| Audio Recording | Planned | Audio capture layer is cross-platform compatible |
| Whisper Transcription | In Development | GPU acceleration requires Windows adaptation; CPU works |
| Ollama (Local + Cloud) | Compatible | Windows installer available. Cloud mode works without local hardware requirements |
| Cloud LLM (Groq/OpenRouter) | Compatible | Requires internet + API key |
| Global Shortcuts | Planned | Alt+Space, Ctrl+Shift+R, etc. |
| Overlay Window | In Development | Window management requires Windows-specific adjustments |
| Auto Updates (OTA) | Planned | Requires Windows installer format (MSI/NSIS) |
| Secure Key Storage | Compatible | Windows Credential Manager |
Known Limitations
- GPU acceleration: Whisper transcription currently uses Apple Metal for GPU acceleration. On Windows, transcription will use CPU until CUDA/DirectML support is implemented.
- Installer format: The Windows installer (.msi or .exe) is in preparation.
- Overlay window: Some window behaviors may differ from macOS.
WebView2 Runtime
Vocoding on Windows requires the WebView2 runtime:
- Windows 11: Pre-installed, no action needed.
- Windows 10: May need to install WebView2 Runtime from Microsoft.
What Is AVX2?
AVX2 (Advanced Vector Extensions 2) is a CPU instruction set required for efficient Whisper transcription. Most x64 processors since 2014 support AVX2.
How to check if your processor supports AVX2:
- Check your processor model at the manufacturer's website (Intel ARK or AMD product pages).
- Common AVX2 processors: Intel Haswell (4th gen, 2013) and newer, AMD Excavator (2015) and newer.
Linux
Linux is not currently supported. There are no immediate plans for a Linux version.