Open source
Read every line it runs.
Privacy claims are cheap; source code isn’t. Typurr is a single Rust binary you can build yourself — and the settings screen has nothing to hide because the config is a JSON file next to your models.
Windows · open source · 100% indoor
Typurr turns speech into clean, finished text at your cursor, in any Windows app — with speech and language models that live on your own hardware, like a well-fed house cat.
That correction wasn’t a typo fix. You said “thursday… actually friday” and it kept only what you meant. (The name, though — the name is a typo. Worked for Google.)
Free and MIT licensed. Runs on your GPU, or CPU-only with Parakeet.
How it works
This is the whole pipeline. There is no fifth step where your voice visits a data center.
Push-to-talk, hands-free toggle, or just say “hey typurr” from across the room.
Whisper large-v3-turbo on your GPU, or Parakeet on CPU with a live transcript while you speak.
A 7B language model (llama.cpp, in-process) strips the “um”s, fixes punctuation, and resolves your self-corrections.
Injected into whatever has focus — editors, browsers, chat, even terminals, safely.
Warm end-to-end latency is ≈1 second on a consumer NVIDIA GPU. Models download once (a few GB) and live in a folder you can see.
The indoor-cat guarantee
bytes of your voice, text, or history. Ever.
There is no server: no account, no telemetry, no audio “retained to improve the service.” Your dictations, your corrections, and everything Typurr learns about how you speak live in plain files on your disk that you can open, read, and delete. Nothing to breach, nothing to subpoena, nothing phoning home in the night.
Features
Everything below is spoken, local, and optional — each one is a toggle in Settings.
Fillers dropped, punctuation added, and mid-sentence changes of mind resolved. Pick how much polish: none, light, medium, high.
“…by thursday — actually friday”Opt-in idle listening that starts hands-free dictation when you call it. It answers to “typer” too — it’s a cat, not a speller.
“hey typurr”Window switching, scrolling, clicking buttons by name (UI Automation), key chords, verbatim typing.
“switch to chrome” “click send” “press control s”It remembers facts you tell it and can search your dictation history. Answers land in a scratchpad and are read aloud with a local voice.
“ask typurr, what did I tell Dana?” “stop talking”Point a local vision model at the monitor you’re on and ask. Nothing is uploaded — the screenshot never leaves RAM on its way to the model.
“what’s on my screen?”Your vocabulary biases the speech model. Your corrections become rules. Your edits teach it style. All stored in a JSON you can audit.
“correct that layla to Rayla”Professional, casual, bullets, prompt-optimizer, or your own custom modes. Say a trigger phrase to expand a snippet. Per-app overrides.
“using professional, …”Capture tasks as you speak and it quietly checks them off when later dictations show they’re done.
“typurr todo ship the report”Get started
Grab the portable zip, unzip, run typurr.exe. Want a Start Menu entry and start-at-login? .\install.ps1 -Autostart does both; -Uninstall removes it and keeps your data.
First run downloads the speech + language models (a few GB) to %APPDATA%\Typurr\models — the overlay names each one and its license while it happens. One time only.
Release, and clean text lands at your cursor. Double-tap for hands-free. Say “what’s on my screen?” or “switch to chrome” when you’re ready for the rest.
Best with an NVIDIA GPU (≥8 GB VRAM, ≈1s latency). CPU-only works for dictation; pair it with Ollama for the full cleanup experience. Prefer compiling it yourself? cargo build --release --features cuda — build docs in the repo. Verify anything with typurr --doctor.
Open source
Privacy claims are cheap; source code isn’t. Typurr is a single Rust binary you can build yourself — and the settings screen has nothing to hide because the config is a JSON file next to your models.