The Windows and Linux betas have received regular testing and are much more stable since the last post. Windows is at about the same feature level as Linux now, including eye tracking. The only exception is I haven’t gotten wav2letter working on Windows yet (there are too many dependencies that don’t work at this time, but I have two promising paths forward that both involve removing most of the dependencies).
Reminder that you can sign up for the beta tier on Patreon to try Talon for Windows (for use with Dragon or for eye tracking) and for Mac/Linux (for wav2letter or eye tracking).
I’ve been working full time on making wav2letter better (the new Talon speech recognition backend I am building). I released a new version a few days ago that has received high praise from the beta testers, including one favorable comparison with Dragon. Many beta users at this point are using wav2letter full time as their main engine. All of my wav2letter models and some information on the training process are posted here: https://talonvoice.com/research/
During this I’ve been working on the “new api”. There are some script repositories and users who are using it full time now. Check out the knausj_talon repository for some examples of what that will look like once the beta is released. As a teaser, here’s a specific example file - newapi comes with a new file format called the “.talon file” which can be used to declare voice commands in an extremely concise manner.
My roadmap right now:
- Further improve wav2letter
- Release a new public version of Talon for Mac with wav2letter
- Solidify the “newapi” overhaul of Talon’s scripting API and release it for Mac
- Polish and release Talon for Linux
- Polish and release Talon for Windows