Why it exists
At sea, speech tools fall apart. Win+H leans on cloud. Whisper stacks often want CUDA and a long install fight. Farting Whisper is a small local answer: Kroko TR-128L via Sherpa-ONNX on CPU — hold-to-talk, type into any focused window.
Hold F11
Hold and speak; release — remaining audio flushes and types into the focused app.
Fully offline
Recognition on your machine. No cloud account for core dictate.
Language packs
Default Turkish 128L; switch EN/DE/FR/… in the UI (download once, then offline).
License
Free for personal use. Do not sell. Full terms on GitHub.
Screenshots
