Why we built it
Most people type at around 40 words a minute and speak at about 150. Dictation should close that gap, but most dictation tools send your voice to a server and charge for it every month. They also handle Urdu badly, and plenty of people in Pakistan think in Urdu and write in English all day. We built Typestop for them first.
What it does
Talk anywhere you can type. Hold Right Ctrl, speak, let go. The text goes in at the cursor in any app, and if a field refuses a paste, Typestop types it in instead.
Runs on your PC. Five speech models ship with the app, from a small fast one to a large one that covers more than 90 languages. There's no account to make and no API key to paste, and dictation keeps working offline.
Urdu in, English out. A second key dictates in a second language, and you can speak Urdu and get English typed.
Cleaner text. Punctuation goes in by itself and filler words come out. A personal dictionary learns names and jargon, and spoken shortcuts expand into whole blocks of text.
Meeting notes. Typestop notices Zoom, Teams, Meet and WhatsApp calls, and no bot joins the call. A local AI model writes up what was decided and who has to do what, and you can ask it about the call afterwards.
Screen recording. Record your screen or a single window with your camera in the corner, then paste the video straight into a chat.
How it is built
The app is Tauri 2 with a Rust core and a React and TypeScript interface. Speech runs through ONNX Runtime and whisper.cpp, on Parakeet and Moonshine models for English and European languages and Whisper for the rest. Silero spots when you start and stop speaking, and Harper fixes the grammar. The installer is 18 MB and needs no admin rights. Updates arrive signed, inside the app. A Cloudflare Worker serves the download site and the update files.
Where it is now
Typestop is in a free private beta on Windows 10 and 11, and it has shipped 32 releases since the first one in June 2026.
