Hi all. I’ve been building TuxTalk, a free dictation app for Linux, and since it’s on the AUR I figured some of you might like it. Hold F8, talk, let go, and the words get typed wherever your cursor is. It’ll transcribe recordings and videos into text or subtitles too. Everything runs on your own machine.
yay -S tuxtalk builds it from source with Arch’s own packages, or yay -S tuxtalk-appimage if you’d rather have the AppImage.
0.22 came out this week and added a Labs option for running things by voice: “open Firefox”, “snap left”, “close window”, “new tab”, “volume down”, and a mouse grid where you say a number or two and then “click”. On KDE it zooms in on the part you pick. It’s off until you turn it on.
Fair warning on how it’s made: Claude writes all of the code, I can’t. I design it, test it in VMs on eight distros, CachyOS included for the AUR packages, and dictate most of what I type with it. The code’s held to what a human developer would expect, readable and tested.
If something breaks on your setup I’d like to hear about it.