1. 19
  1.  

  2. 3

    Ooh! Exciting stuff. I hope somebody figures out an open speech-to-text model in the near future, it would be really neat to have a fully open replacement for those corporate microphones.

    1. 2

      Ooh! Exciting stuff. I hope somebody figures out an open speech-to-text model in the near future, it would be really neat to have a fully open replacement for those corporate microphones.

      There’s always Ada/Almond, as outlined in this (very) old release blog entry:

      https://www.home-assistant.io/blog/2019/11/20/privacy-focused-voice-assistant/

      1. 2

        I agree. In general I’m not a big fan of voice controls, don’t really know why. But home automation is perhaps the one place where I find it very useful. Something open would be amazing, not just to maintain control, but the commercial options require a publicly available server (solved with Home Assistant Cloud, but still).

        1. 1

          I use kaldi-active-grammar on long days (think coding AND chatting with friends a lot). It works pretty well although I do still have to repeat stuff due to using a model that isn’t trained on this microphone and the microphone being pretty bad. It’s much better in the demo video in that README.

          1. 1

            Mozilla speech?

            1. 1

              I haven’t used it for years, but Sphinx worked very well for commands (limited grammar / vocabulary, versus dictation needing to recognise all of English).

              1. 1

                I think you meant to post this on the Amazon Echo thread?

                1. 2

                  No, I’m imagining Home Assistant as the back-end of an open voice assistant someday. Most of what things like the Echo are used for (at least, by me…) is controlling smart devices.

                  1. 4

                    I’d bet on talon, it comes with really really good speech models that you can use right now to write code hands-free, or automate your local machine. Automating a whole home should be a trivial step from there (the main issue I think is feedback), with the right scripts

                    1. 2

                      Oooh, I get it! Yeah, I’d love to ditch the software and keep the hardware of the Echo devices, so I interpreted it as how you would reuse the hardware with stock Linux, but it’s obvious that one wouldn’t need the hardware. :facepalm:

                      1. 1

                        Well, it’s… mathematicians would say “self-evident”. Obvious, but only once you hear it. Easy to miss otherwise. Don’t be too hard on yourself. <3