Capture ideas, write articles, and draft emails at the speed of speech. Completely local voice recognition—dictate instantly even offline.
Optionally plug in the most powerful AI brains (ChatGPT / Gemini / Claude) to auto-polish and format with a single sentence.
Don't let your private conversations become free training data for AI models, and don't let your personal data fuel targeted ads.
Truly permanent, offline, and free—your privacy never leaves your device.
Choose your language:
Join the ChordVox community
Get release updates, ask questions, and share feedback in our Telegram channel:
https://t.me/chordvox6
What you actually said:
"Umm... let's have a meeting this afternoon, maybe at 3, invite Bob... oh wait, I mean invite Alice. We need to discuss the UI redesign."
- 🟢 Standard Dictation: Transcribes exactly that, keeping your filler words and hesitations.
- 🚀 ChordVox AI:
[Meeting Notice] Time: 3:00 PM today; Attendee: Alice; Topic: UI redesign discussion.
💡 A permanently free local speech-to-text engine that turns your spoken thoughts into perfect text.
-
⌨️ Speak into Text, Pasted Instantly — One hotkey triggers the flow: record → transcribe → refine → paste right at your blinking cursor. Zero window switching, unbroken train of thought.
-
🧠 AI Refinement Pipeline — Raw speech → polished prose. Automatically strips filler words and perfects your punctuation. (Power users can plug in their own local GGUF or top-tier APIs.)
-
🎯 Summon Your Executive Assistant — Start your sentence with "Hi ChordVox, draft an email...", and it instantly transforms from a dictation tool into an instruction-following assistant.
-
📖 Learns Your Jargon — A built-in custom dictionary. Feed it your colleagues' names, coding acronyms, or medical terms, and it will transcribe them flawlessly.
-
🔒 100% Local (Free Forever) — Audio processing happens entirely on your machine. No Python setup, no cloud uploads, total privacy.
👀 Click to reveal hardcore geeky specs
- Under the Hood — Built on whisper.cpp, NVIDIA Parakeet, and SenseVoice. Connects to OpenAI compatible endpoints or uses bundled llama.cpp.
- System Hooks — Deep OS integration using AppleScript/Swift (macOS), SendKeys/low-level hooks (Windows), and XTest/ydotool (Linux).
- Dual-Profile Hotkeys — Bind two completely independent hotkey and engine/model workflows.
- Storage Management — 1-click removal of hefty Whisper/GGUF model caches.
| Pain Point | ChordVox Solution |
|---|---|
| Typing is slow; you think faster than you type | Speak naturally → get polished text in < 2 seconds |
| Cloud voice tools send audio to unknown servers | Local STT means audio stays on-device |
| Dictation output is raw and messy | AI refinement fixes grammar, punctuation, and formatting automatically |
| Switching between dictation app and target app breaks flow | Paste-at-cursor removes the copy-paste step entirely |
| Enterprise / medical / legal jargon gets mangled | Custom Dictionary biases the model toward your domain-specific terms |
| You need different AI quality for different tasks | Dual-profile hotkeys: one for fast drafts (Groq), one for polished output (GPT-5 / Claude) |
┌─────────────┐ ┌──────────────────────────┐ ┌─────────────────┐ ┌──────────────┐
│ Hotkey │───▶│ Audio Capture │───▶│ STT Engine │───▶│ AI Refine │───▶ Paste
│ (Globe/Fn/ │ │ MediaRecorder → IPC │ │ whisper.cpp │ │ GPT / Claude│ at
│ Custom) │ │ → temp .wav file │ │ Parakeet │ │ Gemini/Groq │ Cursor
└─────────────┘ └──────────────────────────┘ │ SenseVoice │ │ Local GGUF │
│ Cloud STT │ └──────────────┘
└─────────────────┘
Tech Stack: Electron 36 · React 19 · TypeScript · Vite · Tailwind CSS v4 · shadcn/ui · better-sqlite3 · whisper.cpp · sherpa-onnx (Parakeet) · llama.cpp · FFmpeg (bundled)
After opening the release page, click Show all assets if needed, then choose:
- macOS (Apple Silicon):
ChordVox-*-arm64.dmg - Windows (x64):
ChordVox-Setup-*.exe - Linux (x64):
ChordVox-*-linux-x86_64.AppImage,ChordVox-*-linux-amd64.deb, orChordVox-*-linux-x86_64.rpm
Important
Essential for macOS First Launch: Since this build is not from the App Store, it might be flagged as "damaged" by Gatekeeper. To fix this, run the following command in your terminal:
xattr -dr com.apple.quarantine /Applications/ChordVox.app
open /Applications/ChordVox.appFree forever local offline speech-to-text. Your voice stays private and on-device.
- 🌐 Website chordvox.com
- 📦 All Releases
- 💬 Telegram Community
- 📖 Legacy Technical README
- 📬 Contact:
moonlitpoet@proton.me