Re: Any Progress on an AI Related Screen-Reader?
Hello,
Chime Hart, le jeu. 07 août 2025 07:40:28 -0700, a ecrit:
> Whether its a hook through Speakup or a stand-alone
> project, wondering if there are any commandline tools I can try?
There is piper for instance, which you can try.
> An aspect which might slow down progress may be licensing?
Completely, yes. There is an extensive discussion on the main debian
mailing lists about the question. Such question had already arised in
the case of espeak/espeak-ng, which loads voices as binary data. I had
to package espeakedit to show that these binary data are indeed the
preferred form for making modifications. For AI it's way less clear.
Another aspect is distribution by itself. I see such newer tools
completely ignore good practices in terms of packaging, making
integration into a distribution really hard. I guess we'll have to
embrace the containerization approach to be able to ship such voices
without spending a lot of time trying to make upstream package their
software properly.
Samuel
Reply to: