Next-generation speech recognition running 100% on your computer. 36 languages. Train it on your voice. No cloud. No subscription.
No credit card required · Linux & Windows · Works completely offline
Features
Built for professionals who value privacy, accuracy, and control over their data.
All audio processing happens on your machine. No data is ever sent to any server. No cloud account needed. Perfect for lawyers, doctors, journalists, and anyone handling sensitive information.
See words appear as you speak. NVIDIA GPU acceleration delivers near-instant results. Also works on CPU. Press a shortcut, speak, and text flows into whatever app you're using.
Speak English one sentence, Polish the next. The AI automatically detects and transcribes the correct language. No manual switching. Set favorite languages for faster detection.
Fine-tune the AI model on your voice, accent, and vocabulary. Choose from Quick, Standard, Deep, or Full training presets. The more you use it, the better it gets. Your trained model stays private on your machine.
Add domain-specific terms, names, jargon, or technical words. The AI uses your keyword dictionary to dramatically improve accuracy for specialized vocabulary that standard models miss.
Built-in neural noise removal cleans your audio in real-time before transcription. Background chatter, keyboard sounds, fans — filtered out automatically for cleaner, more accurate results.
Choose the model that fits your hardware. The compact model (1.8 GB) is fast and light. The full model (4.4 GB) delivers maximum accuracy. Both support on-device quantization to reduce memory usage further.
One-time purchase. No monthly fees. No per-minute charges. No usage limits. Speak as much as you want, forever. Perpetual license — the version you buy works forever.
After the one-time model download, the app works without internet forever. On a plane, in a bunker, off the grid, in a secure facility — it just works. No server dependency.
Automatically detects when you start and stop speaking. Filters silence and pauses intelligently.
9 hotkey options including Ctrl+D, Ctrl+Shift+Space, Pause/Break, F13, and more. Hold-to-record or toggle mode.
Use mouse side buttons to trigger recording. Great for hands-free workflows.
Compress models to INT8 precision on your device. Uses ~35% less GPU memory with minimal quality loss.
Select which GPU to use. Runs on dedicated or integrated graphics. Automatic fallback to CPU.
Adjustable threshold to filter background noise before it reaches the AI. Fine-tune sensitivity to your environment.
Review past transcriptions. Optional persistent storage with encryption. Correct mistakes to improve future accuracy.
When the AI gets a word wrong, correct it. Those corrections become training data to make your personal model more accurate.
Modern UI with dark and light mode. Interface translated into 30+ languages.
Browse, switch, rename, and manage your trained models. Instantly swap between base and fine-tuned models.
No GPU? Use optional cloud training tiers to fine-tune your model remotely, then download and run locally.
Text is injected directly into your active application. Browser, email, IDE, terminal, chat — wherever your cursor is.
See It in Action
Designed to stay out of your way. One click to record, one shortcut to transcribe. Every setting exactly where you'd expect it.
One-click recording with live status. See your microphone, AI model, and GPU at a glance. Switch models and languages instantly.
Train the AI on your voice. Choose your base model, name your custom model, and start training with one click. Uses your GPU locally.
Configure hotkeys, recording mode (hold or toggle), volume gate to filter noise, sample rate, and text injection backend.
Set your active language or use auto-detection. Add favorite languages for quick switching from the main window.
Quick, Standard, Deep, or Full fine-tune. Choose the preset that matches your time and hardware. Auto-switch to your new model when done.
All audio processed locally, never saved to disk during normal use. Optional encryption for training data. No data sent to external servers.
How It Works
Download the app (~2.6 GB), run the installer. On first launch, choose your model: compact (1.8 GB) for speed or full (4.4 GB) for maximum accuracy.
Press Ctrl+D (or your custom shortcut / mouse button), speak naturally. The AI transcribes in real-time, streaming text as you talk.
Transcribed text is typed directly into your active application. No copy-paste. Add custom keywords and train the model on your voice for even better accuracy over time.
Languages
Speak any supported language and the AI recognizes it automatically. Switch languages mid-sentence — no configuration needed. Set your favorite languages for faster, more accurate detection.
Why Brethof Voice
| Feature | Brethof Voice Pro | OpenAI Whisper | Google STT | Dragon | Otter.ai |
|---|---|---|---|---|---|
| Privacy (100% local) | ✓ 100% local | ✗ Cloud | ✗ Cloud | ✓ Local | ✗ Cloud |
| One-time price | ✓ Pay once | ✗ Per minute | ✗ Per minute | ✓ $300+ | ✗ Monthly |
| Works offline | ✓ | ✗ | ✗ | ✓ | ✗ |
| Linux support | ✓ Native | ✗ API only | ✗ API only | ✗ | ✗ |
| Languages | 36 | 50+ | 100+ | ~10 | English |
| Auto-detect language | ✓ | ✓ | ✓ | ✗ | ✗ |
| Custom keywords | ✓ | ✗ | ✓ (paid) | ✓ | ✗ |
| Voice model training | ✓ Included | ✗ | ✗ | ✓ ($$$) | ✗ |
| AI noise reduction | ✓ | ✗ | ✓ | ✓ | ✗ |
| Streaming transcription | ✓ | ✗ | ✓ | ✓ | ✓ |
Pricing
Pay once. Use forever. No hidden fees.
14 days, all features unlocked
One-time payment
Per seat, one-time
Download the free trial. No credit card, no sign-up, no cloud account needed.
v1.0 · ~2.6 GB download · Choose your model on first launch: Compact (1.8 GB) or Full (4.4 GB)