Portable archive
Extract anywhere and run <code>./bin/lven-linux</code> — no root, no system install.
lven-instant-linux-0.8.19.tar.gz · 65.7 MB · SHA-256
Lven for Linux
A floating bubble on your desktop. Hold Super+Space to push-to-talk, or click the bubble. Lven types what you said, with the punctuation — it does not rewrite you.
Extract anywhere and run ./bin/lven-linux — no root, no system install. An active subscription is required for transcription.
Version 0.8.19 · 65.7 MB · released Aug 16, 2026
SHA-2562d3c2ca7b7a3f3663ab6ec32b9a13e7f33bc71ebf385fabde81c67721cba7511
Debian package
For Debian and Ubuntu 22.04. Built on 22.04, so apt refuses it on Ubuntu 24.04 or newer — use the AppImage there.
lven-instant_0.8.19_amd64.deb · 49.7 MB · SHA-256
AppImage
Runs on any 64-bit Linux without installing. Mark it executable first (chmod +x). Needs libfuse2, which some distributions no longer install by default.
lven-instant-0.8.19-x86_64.AppImage · 60.3 MB · SHA-256
Updating is manual on every one of these. There is no ClassEve apt repository, so apt upgrade will never offer you a newer Lven, and the app does not update itself — a new version means coming back here and downloading it. The packages carry no GPG signature either; the SHA-256 published beside each file is what proves it came from us.
No runtime to install
The Qt runtime and the on-device speech engine ship inside the archive. What has to come from your distribution is glibc 2.34 or newer — Ubuntu 22.04 and anything later, Debian 12, Fedora 35.
X11, or Wayland with limits
Text goes straight into the focused field through AT-SPI2, which works the same on X11 and Wayland. Where a field does not expose it, the transcript goes to the clipboard and xdotool (X11) or wtype (Wayland) sends the paste. The global hotkey is X11-only — on Wayland, bind lven-linux --toggle in your compositor.
PipeWire or PulseAudio
Standard Linux audio stack. Any microphone recognized by your audio server.
Internet for setup and account checks
The model downloads during setup or repair. Profile, entitlement, and token-refresh checks continue afterward. Transcription itself runs fully on your device, with no microphone-audio upload.