I do like the idea of picking up the phone, and being in contact with the house straight away! It's like sci-fi, but at least 30 year old! "You can just pick up the phone, and tell the house what to do, and it happens" Seems like something that would have been mildly mindblowing at one point in time
If only some clever chap had created an ESP based device which could sense my proximity and be used to automatically listen when I was within a configurable range. Such a device might be named Everything Presence Assist or EP2 perhaps? :) Great video, well done!
Should be doable on the EP1 with a little ESP Home modification. Not something I would do as who knows what it might pickup, but an interesting idea :)
@@EsotericArctos agreed, that's why I'd suggest setting the proximity so that you have to be pretty close to it to start listening, probably also add an LED connected directly to the mic so that you know it's listening. But don't forget, hopefully one day we'll have wakeword detection, then this could be a really cool addition ;)
I have a few extra Alexa Echo's, would be nice to gut some of the components and have the ESP32 take control of the microphone/speakers/LED's to get this working!
I have had success running the voice pipelines in my docker container install. Once I added the whisper and piper container, I just added the Wyoming protocol for each service giving the address and port. I also used self signed certificates to allow the browser mic feature. Not having the esp device or analog phone that was the only way I could go. Great video and thank you for sharing.
Would you mind posting links to the Whisper and Piper docker containers that you used. I see a few of them on my search, but unsure which might work ---Thanks
@@carltonwbrooks Sorry for the delay. I used the docker run command from the home assistant page thread year of the voice chapter 2. YT will not let me put links. But for both Whisper and Piper there is a link in the article for running in docker. I converted these run commands in my docker compose file and then went to integrations and added the wyoming protocol twice. Once for piper and once for whisper giving the local IP address of my docker stack and the Ports the container was running on. I also used mkcert to create self signed certificates so that my browser and phone could use the mic while on the local network. I hope this helps. If you need more detail, I could make a quick and dirty YT video on my channel. Joe
Great Video. Loving the content that you are putting out and the Year of Voice has just made my jaw drop. Would love a video setting up Piper / Whisper on a more robust computer / server while still having all the base home assistant things run on an rpi4 or similar.
I am 8 mins into the video and this seems to have fixed the one annoying issue I had, that I would need to repeat my voice command 6-10 times before google assistant would randomly obey. Google would just keep saying, "Sorry, it looks like the home assistant isn't available right now." and on google home app I could watch my devices would grey out and come back over and over. Imagine that one needs to be pretty desperate to ask home assistant 6-10 times for 1 command to work, eg. light off for bedtime. This is G R E A T!
Imagine having it with a battery pack in form of the communicator from StarTrek. But better would be a wake word detection. But im currently waiting for the amazon echo jailbreak. First impression is that there is just a simple android, which acts as base. Hope its fairly simple to put our own app on it and remove the other cloud stuff. That would be a great base for local voice hardware because it is cheap and easy.
I was configuring it together when I came across another OpenAI configuration in the integrations so I added it from its API and now I added a third Option which is probably Cloud but as an OpenAI conversation agent
Yes, a „real“ alternative to the atom would be appreciated. The phone thing is funny, but I can’t really see anyone walking to a local phone and take a call to switch on some lights.
Let me know how you configure that. I also have asterisk, but currently configured it the autodial way. I'd like to assign home assistant an extension instead but not sure how to do that. Custom trunk?
this is an awesome video! Would love to see another video showing how you got the DIY build using the ESP32 Development Board. I've been trying to modify the ESPHome example code to work with my own DIY build and not having any luck. It seems like assist just hangs in "listening" mode. Thanks for all the awesome content you create for us!
I will be waiting for the option to use this with a voice trigger like, having to press a button kind of defeats the purpose of a voice assistant for me, since I can always use the app on my phone which is close to me most of the time. But it's great to see this progress. BTW Lewis, can I use a microphone and speakers connect to the my HomeAssistant server (I use a desktop as a server), I saw the AI Thinker Audio Kit on Aliexpress it looks interesting would it work?
I have just started to try this. Using an esp home rigged esp32 board with a mic and amp. Still learning the set up but I have clicked a button and spoke to turn off/on a lamp. Side note... I can also send HA tts notifications to the doohickies amp/speaker. Still much to learn & break/unbreak.
Too geeky for me right now. I won't switch from using Alexa to HA for voice until 1/ Magic word is available 2/ There is a simple push button-less microphone device available that is easy to integrate into HA. But I like the ability to turn on/off each voice assistant option for each entity ..8.15 in...(for security and privacy reasons) Very cool. 😀
Thanks again for this very learnfull video Lewis. However I am struggling to install a ssl certificate to get my mic on Android working. Would really appreciatie if you could help out. Assume more less experienced users could also learn from it.
This is so cool! I did the config in my home assistant with a m5stack echo but I can only select english... It doesn't show the others languages. Can someone help me please?
Request: How can we get your video's uploaded to some AI db. I find myself returning to them watching over and over after I get more experience in working through projects, but, still struggling to get them to work. This is at least my 3rd time on this one and I found that answer to many questions that came up since I first watched it. You do a GREAT job in the selection of words you choose to be specific and consistent to both the process and/or the item which you are describing. This makes it easier for AI to ID items in searches. Let me know how I can help. BG
I would love to see a true replacement for the google home/alexa speakers, I rarely use them for audio, I think the esp32 could have great potential here. My vision would be esp32's that could simultaneously act as a voice assistant, multi room audio (with your choice of DSP, used with pre existing audio gear), bluetooth receiver for audio, room assistant, bluetooth proxy and maybe presence detection. Maybe pushing the little beast too far though.
Can't see why it won't happen - Amazon/Google have been basically doing it all for years and all those individual features are being improved in the HA world.
Thanks for the video. I have 1 problem that on mobile app assist voice control works only if you have https connection. Can you create video how to do https with local network? I don't want to open port on my router and i'm using tailscale for remote access. thank you
Thanks for another informative and inspiring video. Can I ask if you plan on making a video about making our own esp32 voice operated system soon? I have all the parts ready to go. Thanks very much.
Finally locally run voice control.... I might start looking more into some smarthouse things then. I was wondering though, do you need to have it push to talk? Cant I set a phrase like with the others, ans say like "Hey house" or something... When we are the point of it being locally run, I don't mind so much it is listening in. With something like a google assistant, I could see the benefit in it only being allowed to listen when I push a button.
I love home assistant, but honestly I was really disappointed when they announced 2023 was going to be the “year of voice” and this update doesn’t seem to change the outlook much It’s not their fault, but they just do not have the resources to do voice *well*. I’m sure this isn’t the final form, but as it stands: - It’s a pain to set up. - The “high quality” voice models still sound incredibly robotic. - You have to talk to it like a robot to be understood, and detection is spotty - It doesn’t really support natural speech - It’s slower - No wake word (also probably going to be impractical to get it working super well with the slow/spotty speech detection) - You have to mostly DIY your own hardware to even use it. They will keep improving things, but not all of those problems can be solved with just more time. They don’t have the massive army of servers to throw at natural language models, or the budget, manufacturing capacity, or volume to put out a smart speaker even remotely competitive with what’s already on the market, especially not at a similar price. I just don’t see this ever being more than an extremely rough imitation of what’s already out there for only the most die-hard of “no cloud, ever” users.
Hi, Lewis. Great video tutorial, like always. By any chance, can you please share your wiring diagram and code for DIY mic and speaker , which you mention in your video. I have esp-32 and amplifier like yours microphone with 3v,gnd,bclk,dout,lrcl and sel pins. Thank you in advanced.
Need a touch-free version. I wonder if wake word will be an option... I'm currently using Alexa with Node-RED to command everything in the home and it's the only thing that works for me and the family / visitors. Not perfect every time, but it's so seamless / ubiquitous and over last few years I've become so used to walking into a room with hands full and setting a lighting scene by number using voice, or turning off garden lights whilst locking the door or arming / disarming alarms or presence sensors. I'd love to move away from Amazon to something local though...
I'm wondering if you could leave that analog phone on call and whenever you need to execute a command you just say for example: "turn on the lights", without any wake word like Alexa or Hey Google.
@Everything Smart Home speaking of languages how can i add Albaninan to it....obviously it is not off the shelf but i am willing to put some work into it..... is there somewhere where people add languages and basic phrases etc ....great video as always
Can this processing all be done on the little home assistant boxes, eg yellow or green? Is it possible / necessary to offload processing to a more powerful box?
Hmm now I wonder if I can do this with my google aiy kit somehow. I had it setup with rhasspy home but even as a satellite it doesn't work the greatest. I am going to have to look into this as that is currently collecting dust since the mycroft project was scrapped.
I'm thinking we have to get a red rotary dial phone now, in which we install an LED that can be triggered to flash, just like in the campy Batman series from the 60s
I’m on a similar idea, but I want to connect it to OpenAI’s integration instead of Assist, and split out the phone audio to mix in with my mic at my desk. That way I can do a bit on a call and people can hear the response.
I found that older AMD CPUs don't support Whisper which is compiled to require SSE 4.1. My old Wyse 7010 thin client has the AMD G-T56N CPU only supporting SSE4a and can't run Whisper. [13:56:18] INFO: Service exited with code 256 (by signal 4) 😥
Just to clarify something in the vid - It is not correct that you can't use the microphone in the browser without a certificate for devices on your local network. There is s setting in Chrome developer settings to enable it for local, insecure, sites.
@EverythingSmartHome I see you have "Norwegian" in your list of languages. Is that only available for subscribers to Nabu Casa? Would be a bit strange since I though whisper did all the magic locally?
Couldn't you wire a Grandstream box to your telephone demarcation box and run all your phones with it? (Does anyone actually own a telephone? How many people had to purchase a landline telephone to do this setup??) Maybe power would be a problem? You could probably run 2 boxes if you rewire the phone jacks or get multi line phones.
Sorry to be negative but, for a relative beginner to Home Assistant like me, setting up devices like the Atom Echo is a nightmare. After many hours of trying, I finally got mine set up without visible log errors. However, I still can't do anything approaching useful with it. In testing, all the pipelines are present and I can switch the Atom's red light on and off from HA. When I press its button, HA detects this but I see no blue or green light as the config says I should and no security notification. You say "if you want to use the microphone option in your browser then you will need HTTPs, but for everything else you don't". As setting up the secure connection has frustrated a lot of us, can you tell us if there's any *other* way to test the Atom? As far as I can see, HA has no interface for using or testing voice *other* than the browser.
I eventually had to give up for the sake of my mental health (after 5 days of effort!). The Atom Echo never worked, and eventually failed to connect altogether, showing encryption errors that persisted after a complete reinstall - a problem the developers say was fixed a month ago. However, I discovered that for me (and contrary to some reports) my phone *can* connect for voice access via the Companion app. Wish I'd known that before I started...
Thanks for the video! Just waiting for the wake word support to ditch Alexa! I also noticed you have the music assistant integration, how did you get it on the newest version of HA?
You mention that multiple pipelines can be set up if you want for example multiple languages. However from what I can see both Piper and Whisper only allow you to configure 1 language. How would a second pipeline be made with a different language?
is there a way to create the pipeline with docker containers? I rul HA in a container on an rpi4 and I want to run wisper and piper on an intel nuc, is it possible?
@@EverythingSmartHome I actually figured that out and just made a quick guide about that on my channel, as it seems to be no tutorial out there so far. I'm sure you or @HomeAutomationGuy would do it better :)
At this stage is the Grandstream the only ATA supported by Home Assistant, or does it support other adaptors, like Sipura or Linksys ATA's ? It doesn't seem like something that would be used outside of being a gimmick, but I am a geek and I like gimmicks lol :)
Great video Lewis. Since I can use my iphone to issue Homeassistant voice commands within the HA app and I always seem to have my iphone (and apple watch) with me, it would be great if it could be made easier by initiate a HA command by for eaxample "raising the phone to talk" or some other simple means. Does anyone know if this is possible?
Great video. I´m missing your promised ESPHome config - also i´m not able to get the replies announced on the M5echo - any idea how to configure that? The speaker does work as audio-player... By the way: my (german) family loves your accent - it´s scottish, correct?
Links added now sorry. Yes you need to add a speaker component to the ESPHome config for the replies, check the ESPHome docs. It is indeed Scottish, good catch! :D
I don't understand the HTTPS requirement of voice. HA is running on a local machine that can be accessed remotely via a Wireguard VPN. Why won't voice work when I access my HA instance from within my LAN? How do I get HTTPS working locally? Can I serve HA via port 443 instead of 8123?
Thanks for the video do you have links to the stuff so I can look into buying/making a speaker? Hopefully soon we won't need to to a push to wake and can just use a wake word like Alexa and Google speakers do
@@EverythingSmartHome Is that the right config for the esp32 DIY version? Appears to be for the m5-stack echo board, not the esp32 dev board. For example, it looked like you were using the boot pin for the wake which is pin 0 vs that config is pin 39 for us on the m5-stack.
I said on the video that I would leave a link to the M5Stack code so you can use it as an example for creating your own one, since there is no way for me to know which pins you will use 😅
Thanks for great video. Just a question. Does the Atom Echo have also wake word capability? If not it seems to be pretty limited as if you need to get out from the couch and go to push a button you might as well use wall mounted control panel.
not using HTTPS even blocks the android app from being able to use the microphone - I wouldn't mind, but it's not like they make it easy to enable HTTPS (having broken my HA install on 4 separate attempts at doing just that) :(
@@EverythingSmartHome i've done them lots (I've worked in IT for 23 years - as a server and infrastructure engineer - so I know my way around a computer) - but not on Linux systems. and when you want to use a cert from your local CA, it becomes almost impenetrable. I took 6 attempts to work out how to setup my Ubunutu Nextcloud to use an internal SSL cert - as none of the information available is actually reliable. and much of it assumes in-depth linux knowledge. I will give it another go - but I am not holding out much hope that I will be able to get it working.
Just look at the Let's Encrypt add-on alongside the NGINX Home Assistant SSL proxy add-on. Should take you all of 10 minutes to set up if you've given it a proper URL. I just have mine go over Cloudflare DNS. Haven't had an issue with it since I set it up years ago.
HA should have called their voice assistant "Home-A", like Homer from the Simpsons. Them we could use a Homer voice for the assistant. It would be epic!
so i just configured 2 M5Stack Atom Echo's just now and i noticed that neither of them show the "controls" box showing the speaker icon but all other boxes are there and it seems to work. am i missing something without that??
I configured mine yesterday and have the same "issue" - like yours it doesn't seem to be affecting anything. Seems to me like it isn't configured as a Media Player - might be fixed in an update
I'm sure it's been said before, but I'm hoping to to get a pipeline from each of the tablet dashboards in my house. That will allow me to complete replace Alexa/Google. Beyond that, with the rise of open source "AI" models, I'd love to see an AI HAOS add-on or integration, so I can pretend that there's a real intelligence behind my home automation.
