95% of the value we get from Alexa is automatic turning on/off of lights and simple functions like cooking timers.<p>We've had Amazon Alexa for five years, since the gen 1 device, and now find it to be an increasing invasion to the sanctity of the home.<p>We find it particularly annoying that nowadays, when you ask Alexa to do something, several times per week it will suggest some annoying upsell crap you don't care about. It used to suggest things once per quarter tops, that was fine.<p>HN, what private alternatives to Alexa may exist?<p>For example, does anyone make a system that's relatively polished and operates entirely in the home, with no audio sent to the cloud? I'd be happy to run a hub/box for the system.
I’m also very interested in an answer. We use Alexa a lot, but this week I had to unplug every Alexa in the house because a distant family member had gained access to the family Amazon account and was trying to use the “drop in” feature to listen to our conversations. In the course of our investigation I found out that Alexa does not log privacy related events at all (there is no record of a drop in stored anywhere), and the UI for locking down which profiles and contacts have access to which feature is unbelievably bad. At this point I can’t prove that this person doesn’t still have access somehow (every single contact taken from your phone has unique permissions to drop in) and I can’t delete the profile this person created without contacting customer support. So the devices are going to stay unplugged until I have time to nuke the Amazon account and create a new one.<p>To be clear, I’m not concerned about cloud processing or even mining my data. Just something I can have a reasonable amount of control over and that doesn’t constantly enable features I don’t want on its own.
It's unfortunate that another comment that links to Rhasspy has been downvoted (I assume because it lacked any other context) so I wanted to mention the project with some additional context: <a href="https://rhasspy.readthedocs.io/" rel="nofollow">https://rhasspy.readthedocs.io/</a><p>While I've not used the entire Rhasspy project myself (but trying it out is on the long list of things to do :) ) I have used the offline Text-To-Speech sub-project Larynx...<p>...and it is <i>amazing</i>!<p>Larynx is <i>significantly</i> ahead in terms of quality of output & variety of voices (fifty--across multiple languages, accents & genders) of any other FLOSS Text-To-Speech project I've tried.<p>I think the relative new-ness of the project is part of the reason Larynx (<a href="https://github.com/rhasspy/larynx/" rel="nofollow">https://github.com/rhasspy/larynx/</a>) currently flies under the radar.<p>If the rest of Rhasspy is as good as Larynx I'd imagine it's worth trying out.<p>Larynx demo video: <a href="https://www.youtube.com/watch?v=hBmhDf8cl0k" rel="nofollow">https://www.youtube.com/watch?v=hBmhDf8cl0k</a><p>Samples of pre-trained voices: <a href="https://rhasspy.github.io/larynx/#en-us" rel="nofollow">https://rhasspy.github.io/larynx/#en-us</a>
I've heard of Mycroft before but have no personal experience with it: <a href="https://mycroft.ai" rel="nofollow">https://mycroft.ai</a><p>It seems like they have an open source client that you can run on your own hardware but it's still dependent on the backend services that Mycroft-the-company provides. Perhaps their privacy stance is more palatable to you, though?<p>I think there used to be a project that was fully on-device but it got bought and consumed by Sonos
I've been wanting pretty bad to help my Google-Home-dependent family back home with this. They mostly need voice-controlled music and lights.<p>First off, unless you want to really DIY the glue-code, you want to use HomeAssistant (huge community) or NodeRED.<p>The only part I'm uncertain about and haven't explored properly myself is the voice-to-text part. Any solution should be pluggable into HA or NR.<p>Relevant threads in HA voice-assistant sub-forum:<p>* <a href="https://community.home-assistant.io/t/replacing-alexa-or-google-home-by-almond-and-ada/152666" rel="nofollow">https://community.home-assistant.io/t/replacing-alexa-or-goo...</a> (Ada, Rhasspy and other FOSS alternatives)<p>* <a href="https://community.home-assistant.io/t/local-voice-control/296830" rel="nofollow">https://community.home-assistant.io/t/local-voice-control/29...</a> (you can apparently use Alexa for voice control even if it does not have any internet access)<p>* <a href="https://community.home-assistant.io/t/best-option-for-local-music-playback-via-home-assistant-and-google-assistant/313652/2" rel="nofollow">https://community.home-assistant.io/t/best-option-for-local-...</a> (local TTS)
Maybe this is an obvious answer, and therefore not one you're interested in, but there's Siri.<p>Although audio is sent to the cloud, you can choose whether it is stored or not, and whatever you pick, Apple's privacy policy is <i>very</i> strict.<p>HomePods are mostly advertised as music playback devices, but I mostly use mine as a HomeKit control device.<p>Homebridge allows you to control non-HomeKit devices via Siri: <a href="https://homebridge.io" rel="nofollow">https://homebridge.io</a>
If you use Home Assistant and want something that is easy to setup, there is Home Intent: <a href="https://homeintent.io" rel="nofollow">https://homeintent.io</a><p>It uses Rhasspy under the covers, and automatically imports lights, switches, fans, etc and sets up sentences and intents for you. After initial setup, it can be used without an active internet connection.<p>All you need is some container knowledge, an extra Pi, and a good speakerphone (like a Jabra Speak 410) to get going.<p>Disclosure: I am the main author of Home Intent.
I've toyed around with mycroft + raspberry + respeaker array. Works well enough that I could get it to figure out wolfram style questions fairly well.<p>Most things seem to assume a raspberry as the base hardware which somewhat implies cloud usage.<p>There is also hive mind which seems to get around the above issue (basically mycroft except with a more satellite mentality to mic placement)<p>Also been wondering whether it is possible to just use the cloud TTS since their voices are quite good. That should fall under GCP/AWS etc terms which seems a little more privacy friendly than straight alexa & friends.<p>Planning on having another go at this, but first tackling presence detection which is a minefield of note too
Although not exactly what you're looking for, I've heard really interesting things about <a href="https://knocki.com/" rel="nofollow">https://knocki.com/</a><p>Essentially you can program your house to coded knocks. Lights/music/door/... all by sequences of knocks.
The harder problem isn't the software - that's solvable to a degree by most of us on this thread - it's the hardware. Alexa et al. are packed full of hardware tweaks such as array based far-field microphones which can determine where the voice commands are coming from, and isolate them from background noise and their own audio output. Then there's the built-in proprietary Alexa/Google-home support in a lot of consumer goods as well. There's hardware widgets that you can get to effectively rebuild Alexa type hardware with an off the shelf SBC, but I think I'm on the ball here when I say most people don't want to do that, and/or don't have the hardware skills either.<p>Personally, I live fine without a voice controlled home assistant. I am however able bodied, and I can press buttons and flick switches without much thought. If that were to change, then yeah, I can see a need for these things, but they <i>really</i> need to be able to work offline with no internet connection should you wish have them configured like that.
This is more of a "DIY" approach but all the tools are there for FOSS and OSHWA solutions.<p>Mozilla has DeepSpeech [0] and, while not as advanced as the stuff from Google or Amazon, my experimentation left me feeling pretty hopeful that it could reliably recognize at least keywords.<p>The Raspberry Pi is quite capable though you'll probably need some dedicated microphone to reliably catch voice data. I know ReSpeaker [1] but maybe some off the shelf conference USB microphones would work as well.<p>[0] <a href="https://github.com/mozilla/DeepSpeech" rel="nofollow">https://github.com/mozilla/DeepSpeech</a><p>[1] <a href="https://wiki.seeedstudio.com/ReSpeaker_4_Mic_Array_for_Raspberry_Pi/" rel="nofollow">https://wiki.seeedstudio.com/ReSpeaker_4_Mic_Array_for_Raspb...</a>
I'm not aware of any private alternatives, but I've been annoyed by the same issue, and since switched to using Siri.<p>So if you've got a few apple devices such as a homepod / apple watch / iPhone, consider the homekit ecosystem, its been working well for me.
I have set up a Rhasspy instance on a Pi 3A (using a Seedstudio microphone HAT) and it works decently, although it is by no means Siri/Alexa grade.<p>The good thing about it is that it takes only a couple of minutes to install via Docker and has all the base niceties out of the box (trigger word, intent detection, speech synthesis).<p>I would say it is good enough to tinker with, although clearly not yet up to par (mind you, I am also trying it in English).
Consider Project Alice. OSS runs on Raspberry Pi or AMD container. <a href="https://github.com/project-alice-assistant/ProjectAlice" rel="nofollow">https://github.com/project-alice-assistant/ProjectAlice</a>
Dumb question: Why does Alexa need to connect to the internet to turn lights on/off or run timers. What happens when it cannot connect to the Amazon mothership.
Polished? No<p>There are open source speech to text engines, text to speech engines and assistant software with APIs, you could probably build something with a raspberry pi, I looked into it a while back but I don't really mind light switches.<p>"Open Assistant" should get the ball rolling for you, search that and dive down the rabbit hole of open source home automation.
I don't want to sound like a smart ass, but I feel like a good alternative to Alexa is just doing all that yourself.<p>I have never had an Alexa but surely there isn't that much time saved by having it turn off lights. Siri on your iPhone can set alarms and shopping lists and presumably your phone it always around you.
If you can get comfortable with Siri (which does send audio to the cloud, though with iOS 15 there's more on-device capability), you should experiment with controlling Google Assistant and Amazon Alexa via Siri.<p>I set up my phone so I can say "hey Siri ok google" and then it asks me what I want to tell google. I then say any command supported by the Google Assistant, and it passes the command through.<p>Technically I think it's supposed to work all in one utterance, but I have found that it never works that way. I always have to split it up. Even still, this is a pretty handy way to be able to ask for information (Siri's knowledge seems quite limited compared to Alexa/Google) without having any always-listening devices in my home.
You could also keep using Alexa and add a do-it-yourself privacy enhancement called Alias: <a href="https://www.instructables.com/Project-Alias/" rel="nofollow">https://www.instructables.com/Project-Alias/</a>
Haven’t seen it mentioned - I think <a href="https://mycroft.ai/" rel="nofollow">https://mycroft.ai/</a> is exactly the sort of thing the original poster as well as the general HN crowd is looking for
mycroft.ai appears to be one option. I heard of them at MakerFaire (2018) and they appear to still be around. I believe they supported installation on a raspberrypi/similar around that timeframe.
I was working on a solution for that a few years ago with an API for raspberry pi<p><a href="https://github.com/victorqribeiro/rpiapi" rel="nofollow">https://github.com/victorqribeiro/rpiapi</a><p>Here you can see I'm using voice commands to drive a car<p><a href="https://github.com/victorqribeiro/raspberryCar" rel="nofollow">https://github.com/victorqribeiro/raspberryCar</a><p>I mean, with a little bit of code and some 3v relays you can achieve what you want
The only viable option that I found that could reliably infer commands from speech is <a href="https://github.com/Picovoice/rhino" rel="nofollow">https://github.com/Picovoice/rhino</a><p>Unfortunately it is not open source (the GitHub just has binary blobs) and requires an account to log in to generate and download model files, but the accuracy is great and you can use it to send commands to Home Assistant to turn lights on/off etc.
One of my back-burner projects is putting some esp32s or similar around the house with microphones, streaming whatever audio they hear to Dragon NaturallySpeaking like in this video <a href="https://www.youtube.com/watch?v=8SkdfdXWYaI#t=9m5s" rel="nofollow">https://www.youtube.com/watch?v=8SkdfdXWYaI#t=9m5s</a> , but hooked up to home assistant scenes / automations instead of emacs commands.
I was wondering the same thing yesterday. It sure would be nice to have something that worked completely on-prem. Would be cool if it could run on something power efficient like an old smartphone. Then you could N old devices, install your app add to wifi, duct tape them to the wall near a power outlet.<p>Would probably be easier with Android as I imagine you could push apk files OTA if you wanted, which would come in handy.
Someone make all offline voice recognition system, you record any catch phrase or “intents” you want the dumb device to recognize as a wake word, no internet needed.<p>This is similar to a feature I saw in waze, you can record your own turn by turn audio directions by recording your voice saying things like “turn left” or “police reported ahead”, it’s how they offer all the fun celebrity and cartoon voices.<p>Thoughts?
I've heard of <a href="https://getleon.ai" rel="nofollow">https://getleon.ai</a> before, not sure if it fits your needs. A quick browse through their docs suggests that you'd have to write a package for the lights, but it doesn't seem <i>that</i> hard at first glance.<p>(It might be harder than it looks especially if your lights' API isn't documented well.)
During school, our AI class had a guest lecture from Monaca Lam talking about Almond [1]. I personally don’t have much interest in virtual assistants, so I haven’t actually tried to use it, but curious to see if anyone else has heard of it or worked on the project.<p>[1] <a href="https://almond.stanford.edu/" rel="nofollow">https://almond.stanford.edu/</a>
Consider Project Alice. Runs on Raspberry Pi or AMD docker. <a href="https://github.com/project-alice-assistant/ProjectAlice" rel="nofollow">https://github.com/project-alice-assistant/ProjectAlice</a>
Does it need to be voice? You can get physical switches that work with “smart” lighting systems, and physical kitchen timers have been around long enough that they are big-free and have a well-tested UX.
I don't have an answer to your question, but have you considered scrapping voice control altogether and instead using motion detectors or other rules to govern whether the lights are on?
Something like Mycroft or Almond(if i recall correctly)<p>Lot of plugins and functionality, especially mycroft.However Almond being open-source is very good for hacking around