CLASSEVE
RouteReleases
Lven Instant / Linux

Speech to text for Linux, on your own machine.

Tap Right Ctrl, speak, tap it again, and the text lands where you were typing. Lven Instant installs as a .deb, an .rpm or an AppImage and transcribes on your CPU — your voice never leaves the box.

Download for Linuxv0.8.41 · 49.6 MB

SHA-256d8bfacb5fff9d6b9ee343051842979775a5a56c3acf050f8033551095f7de5ff.sha256 file

What happens next
ClassEve serversno longer neededYour machine
0.0 MBof 49.6 MBlven-instant_0.8.41_amd64.deb
  1. TransferThe package downloads to your computer. Nothing is installed yet.lven-instant_0.8.41_amd64.deb
  2. Installapt or dnf installs it with a menu entry, and the AppImage runs in place — everything it needs, speech engine included, is already inside.
  3. Speech modelSetup downloads and verifies the model files. This is the last thing on this page that needs a network.
Installing…

Free: make an account, link the app, and dictate.

Three packages · one build

Three ways to install it. One of them fits your setup.

Installv0.8.41
sudo apt install ./lven-instant_0.8.41_amd64.deb
Download lven-instant_0.8.41_amd64.deb

SHA-256d8bfacb5fff9d6b9ee343051842979775a5a56c3acf050f8033551095f7de5ff.sha256 file

.deb49.6 MB
.rpm44.5 MB
.AppImage60.1 MB
Made for
Ubuntu, Debian and their derivatives
Root required
Yes — apt
Touches the system
Installs into the system, with a menu entry
Extra dependency
None — apt pulls in anything missing
Updates arrive by
Download the new package from this page
Integrity
SHA-256 published — verify after download
Why this build

Why Lven Instant on Linux

Linux is usually the platform dictation apps skip. Wispr Flow, Superwhisper, SuperDictate, Whisperstream and Dragon do not ship for it. Lven Instant does, with the same engine, the same account and the same unlimited use as Windows and Android.

The engine is measured, not described. On consumer-grade hardware it runs about six times faster than you speak, 0.93 seconds median per clip, 3.76% word error rate on clean speech. whisper-large-v3 on the same hardware ran 10 to 30 times slower per clip.

One package carries everything the app needs, speech engine included; your distribution only needs glibc 2.34 or newer, PipeWire or PulseAudio, and a microphone it already recognizes. Ubuntu 22.04 and later, Debian 12, Fedora 35 and later.

X11 is the supported session: the app holds the Right Ctrl shortcut and pastes into the focused field itself. On Wayland you bind lven-linux --toggle in your compositor; text goes in through accessibility where the app you type into allows it, and waits on the clipboard where it does not.

Dictate in 25 European languages, detected automatically.

On this deviceWhere your voice is transcribed
six times faster than you speakMeasured speed
3.76% word error rateMeasured accuracy
UnlimitedUsage
about 600 MBMemory
Up to 5Devices on one account
.deb · .rpm · AppImagePackages
FAQ

Common questions.

Which Linux distributions does Lven Instant run on?
Any 64-bit distribution with glibc 2.34 or newer: Ubuntu 22.04 and later, Debian 12, Fedora 35 and later, and their derivatives. PipeWire or PulseAudio for audio; an X11 session for the built-in shortcut.
Does it work on Wayland?
X11 is the supported session, and there the app holds Right Ctrl itself. Wayland does not hand an application a global shortcut, so you bind lven-linux --toggle in your compositor's keyboard settings; text goes in through accessibility where the app you type into allows it, and to the clipboard where it does not.
Is there any other on-device dictation app for Linux?
Open-source Whisper projects exist, and you set up and maintain their models yourself. Lven Instant does that for you: the speech model downloads, checks and repairs itself, the same account covers Windows and Android, and the engine runs 10 to 30 times faster per clip than whisper-large-v3 on the same hardware.
Does it need an internet connection to transcribe?
No. Transcription runs on your own device. Sign-in and the one-time model download use the network; no audio or text ever does.
Is there a daily limit or a word cap?
No. There is no server in the transcription path, so there is nothing to meter. Dictate all day on every device you link.
What does it cost?
Lven Instant is free: make a ClassEve account, link the app, and use it on up to 5 devices.
Compared and explained
The best dictation software for Linux, rankedCompared with OpenWhisprCompared with Wispr FlowCompared with SuperwhisperWhat on-device speech to text isThe speed and accuracy measurement
Requirements

Display server

X11 is the supported session. Wayland does not hand an application a global hotkey, and text goes in through accessibility where the app you type into allows it.

Text into the focused field
Direct insertion
Global hotkey
Right Ctrl, held by the app
When a field refuses
Transcript to the clipboard, app sends the paste

No runtime to install

glibc 2.34+

Everything the app needs, including the on-device speech engine, ships inside the package. What has to come from your distribution is glibc 2.34 or newer — Ubuntu 22.04 and anything later, Debian 12, Fedora 35.

PipeWire or PulseAudio

x86_64

The standard Linux audio stack. Any microphone your audio server already recognizes.

Internet for setup and account checks

no audio

The speech model downloads once at setup. Signing in and access checks use the network; your voice never does.

Also available on
Speech to text for WindowsVoice to text for AndroidLven Cloud for Linux (server-side)Plans and pricing
No ads, no third-party trackers. Sign-in works whatever you choose. We count page views ourselves, without cookies, and while you are signed in they count toward your account. Rejecting stops that, referral tracking and the help chat’s memory of signed-out chats. See our privacy policy.