I love seeing this. I would like to see Mycroft or projects like it being used for more commercial applications.

"Chatterbox is a DIY Kids Smart Speaker that Features Open-Source and Private Voice Assistant, Mycroft"

· Mastodon Share · 1 · 7 · 6

@mike "private" being a relative term. AFAIK, Mycroft is still recording audio and sending it offsite for STT/ASR via DeepSpeech (albeit their own instance). YMMV but I will not have any kind of personal assistant in my home until I can host all components on-prem with nothing leaking to outside services.

@sean If by offsite you mean outside of your own location, then yes. They have their own Deepspeech farm. They're in the process of providing the ability for you to host your own Mycroft back end, but that hasn't been released yet. When it does, you should be able to run your entire house, complete with home automation, with no Internet connection.

@sean It's worth noting also, this particular implementation "Chatterbox" doesn't record anything until a physical button is pushed on a device that you build yourself using off the shelf parts. Really, when it comes to the personal assistant market, it doesn't get more private than that at this point. If not now, in the future, Mycroft could be the standard for privacy in this market.

@mike Yep, I've been sitting on the sidelines watching this situation evolve for some time.

"you should be able to run your entire house, complete with home automation, with no Internet connection."

That's my goal. I wouldn't mind having a Jarvis. But Jarvis lived in Tony Stark's basement, not in some server farm at a cloud provider that data mined everything Tony asked Jarvis to do.

@sean I would love to have a Jarvis, and I think that's where we're headed. Unfortunately we're not all billionaires and Jarvis doesn't run on a $35 appliance. Most people aren't willing to pay for the equipment necessary to make it work, so there's been very little driver in the market to make that infrastructure available. I think Mycroft can change that and provide an open front and back end.

@mike Granted, cost and market size are definitely factors. For running DeepSpeech locally you need a dedicated STT/ASR box with a high-end nVidia GPU to do realtime voice. So the DIY Jarvis would be in the range of $1000-1500. Understandably, most wouldn't pay that for the average HA use case (lights, thermostat, play music). TBH Mycroft would need a more compelling skill base before I'd invest that much myself.

@sean Yes, unfortunately (this is just my opinion) this is the genesis of the voice market, and it's my belief that we're going to see voice in all sorts of places going forward. If there's not an alternative to the big players, then Google and Amazon are going to be used in all these cases. In 5 years you won't be able to microwave a burrito without an alert on your phone telling you the middle is still cold. That's why I'm supporting Mycroft here and now, even though not everything is optimal.

@mike I understand, and agree that is where the puck is moving. I'm not trashing Mycroft, I'm just saying it still isn't private enough for me to consider using it myself. For the masses, the privacy war is over (and the People lost).

@sean I very much hope you're wrong about the privacy war, but you probably aren't. At least if we support alternatives, there could still be a place where those who care about privacy can still get some of the same functionality.

Sign in to participate in the conversation

Fosstodon is a Mastodon instance that is open to anyone who is interested in technology; particularly free & open source software.