Hacker News Reader

Comment by virmundi | Hacker News Reader

I like the idea, but wonder if the same features are possible on a stand alone tool. For example, the device does the NLP there on the premise. It exposes a plugin API for apps. The app gets the text translation by registering a callback URI. Then it can do want ever it wants with that. Expose a streaming receiver api for music. Essential get rid of the middle man service.

Is this possible of a Pi?

View on HN

Possible? Yes. On a Pi? .. Maybe? Probably not with current power I'd guess. Allot goes into the voice training and equivalent operations. There are a few OSS alternatives to build something similar; I don't think anything has a similar power yet though.

There’s a Mozilla project for everything https://hacks.mozilla.org/2018/02/how-to-build-your-own-priv... (voice is experimental and I don’t know if it’s on device)

Looks cool, but I haven’t tried. I’m the kind of person that gets up and walks to the light switch.

There's Mycroft too: https://mycroft.ai/ We have a lot of alternatives but they are unable to get the sort of marketing and hence the attention mainstream devices like Alexa and Google Home do

In fairness ( I'm Mycroft's CEO ) our product isn't on par with Assistant or Alexa....yet. We've had to do a ton of work around wakeword detection, natural language understanding, device management, skills management, user management, credential management and speech synthesis BEFORE we've had a chance to work on the user experience.

That said....we are on track to have a high quality experience when we go to production in Feb 2019 ( Release 19.02 ). As an open source community we'd welcome any help folks have to offer. Contributors don't even need to know how to code, all they need to be able to do is listen to sound samples at https://home.mycroft.ai and tag them as "Hey, Mycroft" or not "Hey, Mycroft"

Thanks to everyone who is helping make Mycroft a reality!

Sure. Have a look at Snips [0], Mycroft [1], Jasper [2] or Adrian [3] which are all open source home assistant devices with voice activation, although some do use Googles NLP cloud APIs for speech processing. Snips seems the most interesting, but looks terrible - a bare PCB on your wall? Really?

0. https://snips.ai/

1. https://mycroft.ai/

2. http://jasperproject.github.io/

3. http://www.theadrianproject.com/

Of course it's possible.

Then all that's left to figure out is...

1. Make it work really well across a vast array of voices, accents, and background noise.

2. Make is easy to add new features to, in order to keep up with the competition.

3. Mass produce it at a price people will pay.

4. Have a viable business model.

Do these simple things and you'll be a very, very rich person.

Yes.

https://snips.ai

Reply on news.ycombinator.com