Oxeegen's voice-to-text desktop app — dictation, meeting transcription, and notes,
powered by Oxeegen's self-hosted AI models out of the box. For macOS, Windows, and Linux.
VOxee turns your voice into text, notes, and actions from your desktop. Press a hotkey, speak, and your words appear at your cursor.
By default it points at Oxeegen's self-hosted, OpenAI-compatible endpoints for transcription and AI — enter the shared organization key once and everything works, with no personal accounts, sign-in, or per-provider API keys. Any scope can be pointed at your own custom OpenAI-compatible endpoint instead.
VOxee is a rebranded, Oxeegen-configured fork of OpenWhispr (MIT). Brand-specific configuration lives under
brand/; upstream changes are merged in periodically.
- No accounts, no BYOK setup — VOxee ships pre-pointed at Oxeegen's endpoints; the user enters a single shared organization API key.
- Per-scope models — transcription, dictation cleanup, dictation agent, translation, note formatting, and chat each resolve to an Oxeegen model.
- Oxeegen or Custom — every model screen offers a locked Oxeegen choice or a Custom OpenAI-compatible endpoint (your own URL, key, and model).
- Voice dictation — global hotkey to dictate into any app with automatic pasting
- Dictation translation — dedicated hotkey to dictate in one language and paste the text in another
- AI agent — a named voice assistant backed by Oxeegen models (or any custom endpoint)
- Voice agent hotkey — sends your dictation straight to the agent as a command — no wake word, no cleanup pass
- Meeting transcription — auto-detect Zoom, Teams, and FaceTime calls, with self-hosted realtime transcription over the OpenAI-compatible
/v1/realtimeWebSocket, an automatic batch fallback (buffered/v1/audio/transcriptions) when realtime isn't available, and a configurable chunk interval - Speaker diarization — on-device voice fingerprinting separates remote participants across meeting audio; your local mic is labelled as you
- Notes — create, organize, and search notes with folders, always-on offline semantic search, and AI actions
- Audio import — transcribe existing audio and video: drag in files, batch-upload, or paste a YouTube/audio URL
- Prompt studio — view, customize, and test the system prompt for each scope; VOxee ships versioned default prompts
- HTTP debug panel — inspect model requests and responses in-app to troubleshoot self-hosted endpoints
- Bring your own endpoint — switch any scope to a custom OpenAI-compatible endpoint; local engines (Whisper, NVIDIA Parakeet, GGUF via llama.cpp) are supported by the underlying app
Deployment config lives in the root .env (see .env.example) and is baked at build time:
VITE_OXEE_ENDPOINT— shared OpenAI-compatible base URL (with/v1)VITE_OXEE_MODEL_*— per-scope Oxeegen model ids
Brand values (product name, app id, theme, feature flags, versioned default prompts) live in brand/config/. Windows packaging must run on a native Windows host.
git clone https://github.com/Oxeegen/VOxee.git
cd VOxee
npm install
npm run devRequires Node.js 24+.
React 19, TypeScript, Tailwind CSS v4, Electron 41, better-sqlite3, whisper.cpp, sherpa-onnx, shadcn/ui.
VOxee is a fork of OpenWhispr, the open-source, privacy-first dictation app. Upstream documentation applies to most core features:
- OpenWhispr docs — quickstart, platform guides, troubleshooting
MIT — inherited from OpenWhispr.
- OpenWhispr — the upstream project VOxee is built on
- OpenAI Whisper — speech recognition model powering local and cloud transcription
- whisper.cpp — high-performance C++ implementation for local processing
- NVIDIA Parakeet — fast multilingual ASR model
- sherpa-onnx — cross-platform ONNX runtime for Parakeet inference
- llama.cpp — local LLM inference for AI text processing
- Electron — cross-platform desktop framework
- React — UI component library
- shadcn/ui — accessible components built on Radix primitives