txscribe for your computer
One app for macOS, Windows and Linux. It can transcribe on the computer itself, so a recording never has to leave it, or through txscribe in the cloud; either way it is the same library, and you can switch in Settings.
No build is published yet.
The desktop app is built; what is missing is a signed release. The buttons will appear here the moment one is published — there is nothing to join or wait for.
macOS
A Mac with Apple silicon or an Intel chip.
Apple silicon
Not published yet
Intel
Not published yet
Windows
Windows 10 or later, 64-bit.
Installer
Not published yet
Linux
The AppImage runs on any 64-bit desktop once it is made executable; the .deb installs on Debian and Ubuntu. Capturing the sound of other apps needs PipeWire.
AppImage
Not published yet
.deb
Not published yet
Transcribing on this computer
When you choose it, the audio is decoded and transcribed on the machine and nothing is sent anywhere. The first run downloads a speech model, a few hundred megabytes, and keeps it. Which model sizes the app offers depends on the computer's memory and whether its graphics chip can be used; a machine without one is limited to the smaller models, which are faster and make more mistakes.
Through txscribe in the cloud, the app uses the same engine as the website — speaker labels, the languages the website lists — and needs an account; signing in happens in your browser and hands the app a key for itself alone.
Checking what you downloaded
Each button above lists its file's SHA-256. The hash of the file you have should match it exactly; if it does not, the download is damaged or is not ours.
macOS shasum -a 256 <file> Windows certutil -hashfile <file> SHA256 Linux sha256sum <file>
Something did not work
Write to support@txscribe.ai with the system you are on and what happened. The support page has the limits we already know about.