daanzu

I develop free, open source, cross-platform software for better controlling computers by voice. Primarily, I'm building <a href="https://github.com/daanzu/kaldi-active-grammar">kaldi-active-grammar</a>: a library that packages and extends the <a href="https://github.com/kaldi-asr/kaldi">Kaldi</a> speech recognition engine for real-time command &amp; control with many complex grammars. I'm using it to write a <a href="https://dragonfly2.readthedocs.io/en/latest/kaldi_engine.html">backend</a> for <a href="https://github.com/dictation-toolbox/dragonfly">Dragonfly</a>/<a href="https://github.com/dictation-toolbox/Caster">Caster</a>, which exemplify this interaction approach. However, kaldi-active-grammar isn't tightly coupled to them, and can be used by other frontends/frameworks as well. 
 
I've been coding entirely by voice for 5+ years, and find major (commercial) offerings sorely lacking. They are designed mostly for generic prose dictation. Their voice commands are limited and verbose, making them frustrating and fatiguing for power users. They support certain languages/dialects/accents, but if you don't sufficiently match them, there's not much you can do about it. They are inflexible, closed systems.
 
 
I want to leverage state-of-the-art speech recognition technology to empower individual users to help themselves as they best know how. If you value this work and want to encourage development, please consider supporting me. Thanks!

Recent posts by daanzu