The FOSS-only crowd might flame me for this, but I’d argue this type of scenario is a legit use-case for voice assistants, because “remind me to buy ________” is a fairly easy habit to get into and it’s a single step, fast enough to beat the attention bounce.
Edit: I meant no offense. Reworded to “FOSS-only.”
Heard some buzz and have been meaning to read up. The speech service has been the primary puzzle piece binding me to proprietary systems, but if it’s time it’s time.
I’ve tried voice assistants, won’t be viable until there’s at least half a million autistic linux users who iron out all the kinks for a self hosted service.
I bet some FOSS voice recognition projects have matured since last I checked, but the closest I had sketched out in the past required external calls to the local speech kit api on macOS or iOS. We’ll get there. It’s too useful to let big tech have a monopoly on it.
We did that for years, until the products we bought switched APIs making us have to change shopping list apps a couple times, then shut off the feature altogether.
Any tech that requires an outside server eventually gets shut off and you spend my more time managing it than you saved in the long run.
Just toss the shampoo bottle on the floor to remind yourself.
Ugh that is the worst! I’ve been using Siri and Shortcuts for these basic commands and automation, which is still all on-device, but ofc they could break my workflows with an update any time just like you described.
So I’ve been migrating everything to my own stack, with the ultimate goal of eliminating all external dependencies, even the voice recognition. That will probably take a few more years at this rate but it’ll be worth it for the self-reliance.
Yes, that is a legitimate use case for that technology.
I do not consider myself anti-tech by any stretch of the imagination (I can put my hands on no less than five computers from where I’m sitting) and I want things like voice assistants and smart houses and whatnot for the benefits they can provide, but we’ve got to pry the invasive corporate bullshit out first.
Yeah I’m with you. In most settings it’s me who is the tinfoil hat. I fully degoogled by 2019, began self-hosting bridgeable services years prior when SBCs and containers made it easier to scale, and all my smarthome artifice is offline save for a limited interface exposed via HomeKit.
But I still make guarded exceptions where the value-add is simply too high to ignore (e.g. using smart phones and fitness trackers) and/or the big-tech privacy commitments still appear to hold (though that’s pretty much down to just Apple now, and I know eventually they’ll turn too).
If it sounds like I want to have my cake and eat it too, that’s because I do, but I agree with you.
Like I stopped using fitness trackers on smart phones because I realized all of them want my data more than they want to be a value add to my smart phone purchase. I don’t want the power company to manage my thermostat because the power company isn’t on my side.
True on both counts. Fitbit is the worst offender I’ve seen so far. You’d think it would be Strava (which I’ve also stopped using outside competitions that require it) but Fitbit is just a health data broker at this point. I aggregate with apple’s health app now. It replaced a bunch of tedious spreadsheets and the watch collects way more useful and granular training data than I ever recorded myself. For PT management it’s just too invaluable to ignore.
Yes tried smart thermostats long ago and they never worked right for me. Nowadays I have zones scheduled mostly by time-of-use rates, sunlight and battery levels, which I think is the same idea behind giving the utility company control of your thermostat (they want to have a chunk of the grid to use for peak attenuation, but time of use pricing accomplishes the same thing without letting them past your firewall).
The FOSS-only crowd might flame me for this, but I’d argue this type of scenario is a legit use-case for voice assistants, because “remind me to buy ________” is a fairly easy habit to get into and it’s a single step, fast enough to beat the attention bounce.
Edit: I meant no offense. Reworded to “FOSS-only.”
It absolutely would be a good thing for a voice assistant.
But most kickback against voice assistants isn’t the lack of use case, it’s all the other bullshit you have to accept alongside it.
If I could install a voice assistant that didn’t require a constant internet connection and could work alongside other services, I’d use it.
I think there’s one on F-droid? Starts with a D. I’m sorry I just woke up and can’t find it.
Edit: its Dicio, oof tho it doesn’t look offline but it does tick the rest of the boxes.
HomeAssistant has seen a bit of a revolution around the concept of voice assistants.
Heard some buzz and have been meaning to read up. The speech service has been the primary puzzle piece binding me to proprietary systems, but if it’s time it’s time.
You can build your own on a Raspberry Pi or similar PC
https://github.com/OpenVoiceOS
Can indeed work offline, but only the basic stuff
This looks promising! I’ll check it out. Really curious about the inference perf on pi.
I’ve tried voice assistants, won’t be viable until there’s at least half a million autistic linux users who iron out all the kinks for a self hosted service.
I bet some FOSS voice recognition projects have matured since last I checked, but the closest I had sketched out in the past required external calls to the local speech kit api on macOS or iOS. We’ll get there. It’s too useful to let big tech have a monopoly on it.
We did that for years, until the products we bought switched APIs making us have to change shopping list apps a couple times, then shut off the feature altogether.
Any tech that requires an outside server eventually gets shut off and you spend my more time managing it than you saved in the long run.
Just toss the shampoo bottle on the floor to remind yourself.
Ugh that is the worst! I’ve been using Siri and Shortcuts for these basic commands and automation, which is still all on-device, but ofc they could break my workflows with an update any time just like you described.
So I’ve been migrating everything to my own stack, with the ultimate goal of eliminating all external dependencies, even the voice recognition. That will probably take a few more years at this rate but it’ll be worth it for the self-reliance.
Yes, that is a legitimate use case for that technology.
I do not consider myself anti-tech by any stretch of the imagination (I can put my hands on no less than five computers from where I’m sitting) and I want things like voice assistants and smart houses and whatnot for the benefits they can provide, but we’ve got to pry the invasive corporate bullshit out first.
Yeah I’m with you. In most settings it’s me who is the tinfoil hat. I fully degoogled by 2019, began self-hosting bridgeable services years prior when SBCs and containers made it easier to scale, and all my smarthome artifice is offline save for a limited interface exposed via HomeKit.
But I still make guarded exceptions where the value-add is simply too high to ignore (e.g. using smart phones and fitness trackers) and/or the big-tech privacy commitments still appear to hold (though that’s pretty much down to just Apple now, and I know eventually they’ll turn too).
If it sounds like I want to have my cake and eat it too, that’s because I do, but I agree with you.
Like I stopped using fitness trackers on smart phones because I realized all of them want my data more than they want to be a value add to my smart phone purchase. I don’t want the power company to manage my thermostat because the power company isn’t on my side.
True on both counts. Fitbit is the worst offender I’ve seen so far. You’d think it would be Strava (which I’ve also stopped using outside competitions that require it) but Fitbit is just a health data broker at this point. I aggregate with apple’s health app now. It replaced a bunch of tedious spreadsheets and the watch collects way more useful and granular training data than I ever recorded myself. For PT management it’s just too invaluable to ignore.
Yes tried smart thermostats long ago and they never worked right for me. Nowadays I have zones scheduled mostly by time-of-use rates, sunlight and battery levels, which I think is the same idea behind giving the utility company control of your thermostat (they want to have a chunk of the grid to use for peak attenuation, but time of use pricing accomplishes the same thing without letting them past your firewall).