Real-time speech-to-text for Windows, macOS, and Linux. Whisper runs locally. Optional AI polish runs locally. Your audio never crosses the loopback.
Most dictation tools ship your microphone to someone else's server. Lokut doesn't. The transcription engine and the polishing language model both run on your CPU or GPU. No account. No telemetry. No phoning home.
Whisper transcribes locally. Audio is captured, processed, and discarded. The mic stops at the loopback interface.
Windows, macOS, and Linux (including Wayland and X11). One purchase covers all three. Rare in the genre.
An on-device language model trims fillers, fixes punctuation, normalizes tone. Optional, never required.
$10 or $50 — once. The version you buy runs forever. Update renewals optional. No subscription, ever.
If your hardware can't run the local LLM well, paste an Anthropic, OpenAI, or Groq key — direct, no middle service.
Older versions become open source on a rolling 3-year delay. Trust by guarantee, not by claim.
No subscription. One-time purchase. The version you bought keeps working — renewals are optional, not punitive.