Complete Privacy
Stenox can operate 100% locally on your Mac. When using local providers, your audio and text never leave your device — perfect for sensitive content, confidential work, or anywhere privacy matters.Local Providers
WhisperKit
Local TranscriptionApple’s optimized Whisper implementation. Runs on your Mac’s Neural Engine for fast, private speech-to-text.
MLX
Local AI EnhancementApple’s ML framework running efficient language models. Grammar correction and formatting without cloud.
Why Go Local?
Complete Confidentiality
Healthcare, legal, financial, or any sensitive content stays on your Mac.
Offline Functionality
Works on airplanes, in remote locations, or when internet is unavailable.
No Third Parties
Your data is never sent to external services or servers.
No API Costs
No per-minute charges or API fees. Processing happens on your Mac.
Quick Setup
1
Open Stenox Settings
Click the Stenox menu bar icon → Settings
2
Configure Transcription
Go to Models tab → Set Transcription Provider to WhisperKit → Download a model (base.en recommended)
3
Configure AI Enhancement
In the same tab → Set AI Enhancement Provider to MLX → Download a model (Gemma 3n 2B recommended)
4
Verify Offline
Turn off Wi-Fi → Test a dictation → It works!
Requirements
MLX requires Apple Silicon. Intel Mac users can still use WhisperKit for local transcription, but must use cloud providers for AI enhancement.
Model Options
WhisperKit (Transcription)
| Model | Size | Speed | Best For |
|---|---|---|---|
| base.en | 142 MB | Fast | English only, quick setup |
| small.en | 483 MB | Medium | English, better accuracy |
| base | 142 MB | Fast | Multilingual (100+ languages) |
| small | 483 MB | Medium | Multilingual, better accuracy |
MLX (AI Enhancement)
| Model | Size | Speed | Best For |
|---|---|---|---|
| Qwen 2.5 1.5B | ~1 GB | Fastest | Quick tasks, 8GB Macs |
| Gemma 3n 2B | ~1.2 GB | Fast | Best accuracy, recommended |
| Phi-3 Mini | ~2.4 GB | Medium | Balanced quality |
| Gemma 3n 4B | ~2.5 GB | Slower | Higher quality |
Performance
Processing times with local providers:| Configuration | Total Time |
|---|---|
| WhisperKit base.en only | ~3-5 seconds |
| WhisperKit + MLX Qwen 1.5B | ~5-7 seconds |
| WhisperKit small.en + MLX Phi-3 | ~10-12 seconds |
Local processing is slower than cloud (which can be ~1-2 seconds), but provides complete privacy.
Privacy Use Cases
Healthcare & Medical
Healthcare & Medical
HIPAA compliance, patient confidentiality. Use with EMR/EHR applications and custom medical vocabulary.
Legal Work
Legal Work
Attorney-client privilege, confidential documents. Add legal terminology to vocabulary.
Financial Services
Financial Services
Client confidentiality, regulatory compliance. Works offline for secure environments.
Journalism
Journalism
Source protection, sensitive investigations. No data leaves your device.
Personal Privacy
Personal Privacy
Personal preference, data sovereignty. Complete control over your data.
Hybrid Approach
Want privacy for audio but cloud quality for text enhancement?- Transcription: WhisperKit (local) — Audio stays private
- Enhancement: Google Gemini or Groq (cloud) — Text enhanced with top-tier models