Get started
One click. Then just speak.
About a second from voice to text, at any dictation length, and it keeps none of what you say. No voice logs, no training on your words, every dictation runs through TypeMic’s own Predator engine, engineered in Rust, privacy by design. Download, install, and dictate, the text lands at your cursor in terminals, browsers, IDEs, documents, and chat.
Ubuntu · Debian · Mint · Pop!_OS
RecommendedDouble-click the file after it downloads. Your system’s software installer opens, asks for your password once, and sets everything up, permissions, service, and app icon included. No terminal, no manual steps.
Download TypeMic (.deb)64-bit · ~335 MB · one native Rust binary, signed · sha256
Prefer the terminal?
Same install, one line, any Debian-based distro:
$ curl -fsSL https://typemic.com/install | bashFedora, openSUSE, and Arch packages are on the way, tell us which one you need and we’ll prioritize it.
On a Mac? Download TypeMic for macOS instead, a universal build for Apple Silicon and Intel, ~4 MB, signed and notarized by Apple (.dmg).
Speech recognition model: “Parakeet TDT 0.6B v2” by NVIDIA, licensed CC-BY-4.0. Converted to GGML format and quantized (q4_0) by Raptor Labs; the weights are otherwise unmodified. Releases are signed with the RaptorLabs release key.
Then just speak
Hold your shortcut, Right Ctrl by default, speak, and release. Your words appear at the cursor: terminal, browser, IDE, document, or chat. Also on macOS; Windows is planned, join that early-access list.