A Raspberry Pi kicking around one’s workbench is a project waiting to happen — if they remain unused long enough to be considered a ‘spare.’ If you find you’ve been pining after an Alexa or your own personal J.A.R.V.I.S., [Novaspirit Tech] might be able to help you out — provided you have a USB mic and speaker handy — with an accessible tutorial for setting up Google Assistant on your Pi.
A quick run-through on enabling a fresh API client on Google’s cloud platform, [Novaspirit] jumps over to the Raspbian console to start updating Python and a few other dependencies. Note: this is being conducted in the latest version of Raspbian, so be sure to update before you get underway with all of your sudos.
Once [Novaspirit] gets that sorted, he sets up an environment to run Google Assistant on the Pi, authenticates the process, and gets it running after offering a couple troubleshooting tips. [Novaspirit] has plans to expand on this further in the near future with some home automation implementation, but this is a great jumping-off point if you’ve been looking for a way to break into some high-tech home deliciousness — or something more stripped-down — for yourself. Check out the video version of the tutorial after the break if you like watching videos of guys typing away at the command line.
18 thoughts on “Sudo Google Assistant”
make me a sandwich
sudo make me a sandwich
My wife removed me from /etc/sudoers
username is not in the sudoers file. This incident will be reported.
Reported to who, exactly?
Santa Clause, of course!
Just a note of caution, use a secondary google account for this or live with the consequences of some of the security settings you may need to relax. Otherwise It is a very cool project to play with.
I wouldn’t even want to visit a house, where Google listens all the time.
Yea, what is the equivalent to glasshole for this listening tech?
I would suggest the equivalent is “it’s my home go away if you don’t like it”
Seriously though, I would much prefer to have a local speech recognition process than something remote. Not least to be more secure, but also to be able to cope with occasional internet outages.
There are two ‘arguments’ companies make against this: One, it’s more efficient to run it centralized, rather than having a high-power device running processing at home. It’s not an entirely unreasonably a viewpoint, but speaking idealistically, the Xbox and my desktop PCs are examples of powerful machines with remote wake-up and job scheduling capability; it would be nice if we could design technology to interoperate like that to allow the network to temporarily wake a device, use it for processing speech inputs (identifying ‘ok google’ is relatively easy, compared to translating and acting on random speech – you wouldn’t need high power devices running constantly). I also have the Raspi running my print server, the raspi running my file server, the raspi acting as a SIP phone, the Raspi serving as a network gateway and the raspi running my roomie’s TV. Then there are two APs on the internal network with decent applications processors, a hardware SIP phone – and all these are ARM devices with similar instruction sets. I wonder how parallelizable the speech recognition task is, whether it could be designed to run on a ‘cluster’ of spare compute power?
The second argument is intellectual property; when your speech services are hidden in a secure compute space you control, it’s much harder for hackers, competitors and those engaged in international industrial espionage to get a look at its inner workings and undercut you.
Personally, I’d love a permanent ‘agent’ or ‘ancilla’ that ran only using processing I control… but I don’t see how that’s possible until enough of the technology is available under some sort of free licence; and I don’t see that happening at all in the next fifty years.
If I could buy/build an appliance that did all the speech recognition processing locally, I’d do it in a heartbeat! You’re right, though, the backend stuff won’t be available for decades. Sad!
I’d like to see one of these projects that doesn’t even need to use a cloud server or one where you setup you own.
Most of that back end stuff runs on commodity rack mount servers so an old PC should be up to the task.
Though I doubt Google and Amazon would ever release the code to their server side stuff.
Does the Google Assistant implement purchases of some kind?
OK Google, buy a fidget spinner.
So any house with an android phone?
Or a PC
This all brings me back to the Google Whopper fiasco a few weeks ago. I tried to explain to my girlfriend why it was such a big deal, and showed her the news coverage. Like, she didn’t see that the commercial was triggering the response from the devices (because Google had already fixed this by the time).
She said, “I don’t get it! What’s the big deal about the commercial saying ‘Okay, Google, what is a Whopper?'”
*BLING* A WHOPPER IS A HAMBUGER CONSISTING OF…
Fiasco? That was an absolutely brilliant marketing stunt! Because, well, you’ve heard about it, right? Every techie knows everything about this low cost but extremely high visibility commercial! You probably remember the entire commercial? And you probably even know what’s in the BK Whoppers now, right? And Google and Wikipedia responded publicly, adding even more value to the marketing!
Like or or not, this was one brilliant commercial!
Please be kind and respectful to help make the comments section excellent. (Comment Policy)