Download free

download

Grab it, press Ctrl ` , talk.

No account, no license wall, no word caps. One command — or one download — and a one-time model fetch; then speech to text works with the network cable pulled out.

Windows

v1.4.0 · x64 · Windows 10/11

install with one command

Open a terminal ( Win X , then Terminal or Windows PowerShell), paste the line, press Enter. No browser download, no “keep this file?” prompts — and the same tool brings every update.

winget recommended

Built into Windows 10/11 — one command installs the regular app, and winget upgrade keeps it current.

winget install Croppick.CroppickVoice

Scoop

Installs the portable build from our own bucket — and scoop update croppickvoice picks up each new release.

scoop bucket add croppick https://github.com/EvaldasVasiliauskas/scoop-croppick
scoop install croppick/croppickvoice

direct download

Prefer a file? The installer is digitally signed — Windows names the publisher (MB Croppick). While the app is still new, your browser or Windows may add that the file “isn’t commonly downloaded”: that’s a popularity check, not a verdict on the file — choose Keep (or “More info” → “Run anyway”) and carry on.

Installer (.exe) Portable .zip — extract & run
Size
~65 MB + the speech model of your choice (downloaded once, cached)
checksum · installer
published in latest.json
checksum · portable
published in latest.json

macOS

in validation

the app is built cross-platform and already runs from source; the packaged macOS build is being validated. Apple Silicon (M-series) · asks for the usual Microphone, Screen Recording and Accessibility permissions. It’ll appear right here the moment it’s ready.

Linux

in validation

the app is built cross-platform and already runs from source; the packaged Linux build is being validated. X11 today; portable tar.gz and .deb · Wayland capture in progress. It’ll appear right here the moment it’s ready.

first run

Ninety seconds to first dictation

  1. 1

    Install it

    Paste the one-liner (winget or Scoop) or run the signed installer — a minute either way. In-app updates are digitally signed and verified before they install, and any file can be checked against the checksums above.

  2. 2

    Let the model land

    On first launch the default speech model downloads once — the floating button shows live progress. After that, transcription works fully offline; other models from the catalog download on demand if you switch.

  3. 3

    Press and speak

    That’s it. Ctrl+` toggles dictation, Alt+` starts Crop & Speak, Alt+1 captures without voice. All rebindable in Settings.

Privacy at a glance: your audio and screenshots never leave the machine — the full network-use table.