OmniScribe is a private, offline-capable voice-to-text dictation app that works in any app. Speak naturally — your words appear wherever you're typing. No cloud, no telemetry, no compromises.
Paste this in your terminal, or download OmniScribe.
Works in browsers, terminals, IDEs, Slack, email — any text field on your system. Not locked to one editor.
Speech-to-text and AI polish run locally. Your voice never leaves your device. No cloud, no telemetry. We couldn't listen even if we wanted to.
Local inference means words appear as fast as you speak. No round-trip to a server. No waiting for transcription.
Pro tier runs with zero internet. Free tier syncs only a word counter every 5 hours — no audio, no text, just a number.
Voice mode cleans grammar. Notes mode structures into Markdown. Email mode formats professionally. All local, all instant.
Use our built-in models or bring your own OpenAI, Anthropic, or local LLM keys. You control what processes your text.
Not locked to one editor. Dictate into any text field.
Everything runs on your machine. Voice never leaves your device. We couldn't listen even if we wanted to.
Pro tier needs zero internet. Free syncs only a word counter every 5 hours. No audio, no text, just a number.
No cloud round-trip. Speech-to-text runs on your GPU or CPU. Words appear as fast as you speak them.
Everything in Free. Pro adds unlimited words and full offline mode.
All processing is local and instant.
Use your own OpenAI, Anthropic, or local LLM keys. You control which model processes your text.
Set your own key combos for Voice, Notes, and Email modes. Switch instantly without opening the dashboard.
Hotkey, microphone, VAD threshold, model size. One TOML file. Calibration wizard tunes for your voice.
Translucent pill lives above every window. Shows mode, word count, and recording status.
Sign in, upgrade, switch modes, view usage. All from a terminal. No browser needed.
$0/ forever
$5/ month
Enterprise (team management, SSO, admin) and Windows & macOS coming soon.
OmniScribe uses Whisper models, the same speech recognition behind major transcription services. The calibration wizard tunes detection for your voice and microphone. Larger STT models available for higher accuracy. Calibration guide
Yes. Processing is 100% local. Your voice never leaves your device. Free tier syncs only a word counter (a number, not text). We are a subscription business. We don't collect, sell, or profile data. Privacy policy
Sign up activates the app and enforces the free tier word cap fairly. Use Google, GitHub, Microsoft, or email. 30 seconds. No credit card for Free.
Only your word count per 5-hour window. A single number. No audio, no transcripts, no keystrokes, no app usage. We can't see what you dictate.
Yes. BYO-LLM on all tiers. Bring OpenAI, Anthropic, or local LLM keys. You control model, cost, and data. We never see your LLM requests.
Yes. X11 and Wayland both supported via ydotool. Some Wayland compositors restrict global hotkeys. Try X11 if hotkeys don't register.
We sell Pro subscriptions. $5/month for unlimited words and offline. No ads, no data sales, no VC. Our incentives align with yours.
Private. Offline on Pro. Free to start.
Download