Local voice dictation for Windows with intelligent text processing. Push-to-talk with Pause, Whisper STT offline — text inserted right at your cursor.
Three steps. No cloud. No subscription. No nonsense.
Hold Pause from anywhere on Windows. The microphone activates instantly — VOCIX lives in the system tray and is always ready.
Talk normally. Release Pause when done. Whisper transcribes your speech locally — nothing ever leaves your machine.
Processed text appears right at your cursor, in any application — email, browser, IDE, document. No paste required.
Choose how your speech gets transformed before it hits the page. Switch modes instantly with keyboard shortcuts.
Strip filler words and light corrections — preserves your meaning, removes "um", "uh", false starts. Runs 100% locally, zero latency.
Rewrites your speech into professional, formal language suitable for email, reports, or client communication. Powered by Claude.
De-escalates aggressive or frustrated speech into polite, firm communication. Preserve meaning, remove heat. Powered by Claude.
No cloud lock-in, no monthly fee, no account required. VOCIX respects your privacy and your wallet.
faster-whisper runs fully on your machine. Your voice never leaves your computer. No API key, no internet, no privacy concerns. Model downloads once on first launch.
Text is inserted at the cursor position in any Windows application — email clients, browsers, IDEs, Office apps, chat tools.
Speak any of ~50 Whisper-supported languages. Optional one-click offline translation to English — no API key needed. Full DE/EN UI switchable at runtime.
The last 20 dictations are saved in the tray — click to re-insert. Define your own shortcuts like /sig or /adr that expand to full text before insertion.
Business and Rage modes run on the LLM you choose: Anthropic Claude, any OpenAI-compatible API (OpenAI, Groq, OpenRouter, LM Studio …) or local Ollama models. Per-mode override — cloud where it matters, local where it counts.
Choose your Whisper model (tiny → large-v3-turbo). GPU mode auto-detects CUDA — or force CPU/GPU via config. Speed vs. accuracy, your call.
Install via winget install RTF22.VOCIX, Scoop, or just unzip and run — no admin rights, no registry. Auto-updates from the tray with SHA256 verification.
VOCIX is MIT-licensed and lives on GitHub. Fork it, extend it, contribute to it — or just use it. No strings attached.