Gotta say, the way you step through these instructions very clearly, with no over the top quirky personality, AND showing multiple issues you can encounter and how to resolve them, is a breath of fresh air in the smart home automation space. Keep up the good work, I'm subscribing to see more!
I do believe you are one of the best if not the best UA-cam Tutorial Content Creators I have ever viewed and subscribed to and trust me, my subscription list is very exclusive and normally has to be earned over many videos and seldom is it earned on the first watch. I don't mean to embarrass you but I'll even clarify a few reasons why (for the benefit of other creators that could do worse than take a few tips from here). Verbally, you're easy to understand, clear and concise, I don't have to rewind and replay to hear what you said. You use visuals like highlighting the commands and text that you are talking about. You explain what the command is actually doing which is useful education for people that have blindly used similar commands before but had no idea what they were actually doing AND importantly, these descriptions are just the right length, not to in-depth that lead down a rabbit hole. Finally, you haven't jumped on the band wagon of coming up with a cringy slogan or start every video with an emphasised chant like "let's dive right in" etc. A huge thanks for being more real!! I predict your channel and subscriptions growing fast and exponentially and deservingly so!
I agree with all this and subscribed on day 1 after building my wyoming board on a pi 4 with USB Mic. I'll add that your github written tutorial is excellent as well. Thanks for taking the time!
Thanks for the tutorial. You don't have to type "y" if want to proceed. If [Y/n] just press Enter for Y, if [y/N] just press Enter for N. Default choice is always a capital letter.
We'll have to see how it holds up when things inevitably change in a few months, but for now, this is perhaps the best, easiest true step-by-step tutorial I've ever followed on UA-cam. I even followed along for things I already know how to do, like exiting nano.
Followed these instructions and got an identical config up and running with a Pi Zero 2 W and ReSpeaker 2-Mics Pi HAT. For those in the UK I picked up the ReSpeaker at Mouser. Thanks for the clear step by step instructions, easiest hardware/software project in a while.
I had an interesting issue trying get this working but was able to figure it out. I already had followed your ChatGPT instructions and had OpenWakeWord, Piper and Whisper installed. For some reason Home Assistant wouldn't find my new Satellite. I was able to find the instructions to add it manually (using the IP address and port 10700) and then was able to successfully complete the process. I had a set of headphones plugged into the satellite and I'm using the same wake word as your video. Before I installed Open Wake Word locally, it was using what I had already installed in Home Assistant. This is Brilliant! Now to make some changes so that the output of my conversations are redirected to the in-room Sonos speakers!
This is awesome. I've been seeking out a whole house local voice assistant solution and this seems to be pretty close to what I desire. Thanks for the inspiration.
Hello from Norway! Thank you sooo much for this. This is brilliant and works like a charm. Keep up the great work - I really appreciate these awesome videos of yours! -Best regards Kip
I'm actually moving in the next month and wanted to get away "she who shall not be named." Your videos are going to be my roadmap to a whole new smart home setup. Thank You!!!!!
Love it. Run a small test with the product make sure it suites your needs then scale up to support the entire home. Not trying to be too pushy, but if you need help along your journey I do consulting. ;) Have fun @cliffordraymond9123!
Great tutorial. Ordered parts using your affiliate links - I'm gonna try this out myself. This is much more practical than using an M5Stack Atom Echo, which I've already done. You've got a new subscriber!
I've been bothered that out of the box this tutorial (the wyoming-satellite, not your vid) doesn't ship with the ability to do tts notifications. A simple install of mpd solves this nicely - making the communication feel two way rather than just it reply to the wake word. Good job on the vid! Keep making them.
Thanks for the good video vibes. Yea, you’re right. People definitely want to use this hardware as a media player. FR has been submitted to the Wyoming project here: github.com/rhasspy/wyoming-satellite/issues/40 I’m gonna try installing MPD on the Pi soon so see if I can unlock this feature. Should work.
I'm creating a bunch of these around the house, with the eventual plan to repurpose some old Alexa Dots (and some day replacing the rPI with a custom ESP32 board). I control them all remotely via Ansible, which makes life a LOT easier for managing the inventory.
@@FutureProofHomes is it possible to use Echos with this approach instead of replacing (or supplementing) them with these Wyoming devices? I'm certainly going to get that question from the family.
Okay, I got this setup working on a Raspberry Pi 4, and have it partly configured on two Pi Zero 2 W boards. I already had one ReSpeaker, but I'm waiting on the other two I ordered. Now to take it to the next level and get it working with PulseAudio and Snapcast as a music player.
@@FutureProofHomes Can you give me the img file after successfully installing all the steps? I can't install it forever, it keeps giving me an error. thank you.
This is the coolest shit ever! Im pretty tech savy and love learning and just started looking at building a home server and home assistant. Im a big user of google and wifi home lights etc but have always dreamed of it being local and always worried about google listening to me lol i love learning and will attempt this tyvm! ❤
Great job on the tutorial. Disappointing that this level of configuration is needed for HomeAssistant projects. Yes, it makes it the most flexible home assistant ever..... with an ever-increasing learning curve. Why not just release a SD card image and make a few updates for your settings? Some of us just don't have the time to be HA certified.
I'm loving your tutorials and everything you are doing. I think you are on the forefront of the future of the smart home and its damn exciting to watch and build alongside you. I've been a HA fan for 7-8 years. I built your FastGPT pipeline last week, pretty fun to use but so far only light usage. I'm curious if you have thoughts on the approximate cost in monthly OpenAI bills if used that FastGPT pipeline heavily to run your smart home, replace the voice assistant and field general queries. I know it's highly dependent on usage obviously, but wondering your thoughts on a ballpark of that.
Is there any way to get a little screen on the satellite to display a home assistant dashboard or text output of the assist conversation? Trying to replace my Google Hub functionality.
Built this and it works like a charm, but you're gonna have to dub out wakewords in your videos from now on as it apparently really likes your voice... 😆
Thanks for the great guide on how to build a Voice Assistant at home. I am currently planning such an installation in my new home and am wondering about the best solutions. What do you think about other Respeaker boards? for example the "ReSpeaker 4-Mic Array for Raspberry Pi" or the "ReSpeaker 6-Mic Circular Array Kit for Raspberry Pi". I want the best possible quality of microphones so that I don't have problems with the wake word and that the board reads commands well. Can these boards be integrated into the HA in an analogous way? Is this a good idea or is it a waste to overpay?
Honestly, the 2-mic Pi hat has served my needs. The 4-mic pi hat is probably better though. Let me know your results! The 6-mic circular array may not work with the drivers I fear, but try it! That all said, I would still recommend the 2-mic hat. It really is comparable to the Alexa/Google based on all my testing. Even works with background noise in a big room if your raise your voice a bit (which I had to do with Alexa as well).
how come nobody had an issue with installing pysliero-vad? I get the error, no matter what: ERROR: Could not find a version that satisfies the requirement onnxruntime=1.10.0 (from pysilero-vad) (from versions: none) ERROR: No matching distribution found for onnxruntime=1.10.0
Great tutorial! especially for the visual, makes it easier to understand ---- can you do a video on maybe utilizing the GPIO/12 port for external lighting cue? be nice to have another indicator avenue if we are unable to see the 2-Hat
Thumbs-up on the video, but I have a small question: Why did you go with "legacy 64-bit" rather than the current 64-bit OS? From what I read, the current lite version does fine on the 2w, but I figured you must have a reason for going with the legacy version.
Bookworm (newest raspi OS) is still rough around the edges and introduces some breaking changes to python env. Just went through safe route to avoid unnecessary headaches.
Another amazing video. I'm in the process of migrating from a heavy Alexa automation to Home Assistant and this solves my Echo device issue. I have a few questions - 1) I have 12 Echo devices and was wondering if I can set up one of the Wyoming Voice Satellite's and then copy that SD for each of the other RPi's and change the name/user info to so that I'm not having to follow these steps on each device? 2) Is there a method to determine the person speaking? I would like to include a response specific to that person 3) In my home automation setup I have Sonos is each room - would it be possible to have each RPi output the audio to the Sonos device in the room where he RPi is via Home Assistant? I agree with your approach on muting the music to allow the voice conversation to happen - that is the behavior I have now with Sonos in the room and the Echo devices. I would love to be able to get the conversation over my Sonos devices in each room. Keep the great content coming!
1. YES! You can do this. If/when you do, please document the process and share. 2. Give each person their own wake word. I use "hey bob" and you use "hey sally". That's the best way today. 3. My most recent video tries to get rid of Sonos completely. But I also want what you're describing! Yes, this should be possible. Watch my new video and modify the "--snd-command" so that it PulseAudio will use the Airplay protocol to stream to your Sonos. That should be it.
Great video, allowed me to get started with this very quickly. At the moment I can't get the LED setting to work, if I enable it then the "awake" is triggered but the unit then just hangs and needs restarting. Have you ever had an issue with the LED? Also, couldn't get "ok nabu" to work but "hey jarvis" is fine - must be my pronunciation! (Having issues with the ChatGPT pipeline but need to watch you video again to re-set that up - I did it sometime before I got the hardware.)
I appreciate the cool content! Thank you for the inspiration, and keep doing great work! I've just subscribed to your channel :). Now I have to figure out how to get this thing running in my native language :))
This is awesome! Not having an open mic to the internet is something I have been looking forward to (and so has my wife). About the only thing missing as far as I can tell is the capability to use a custom created wakeword. My programming skills are limited; if it is possible to utilize a custom wakeword please let me know.
This has been so helpful! Thanks! Where do we go to figure out how to evaluate the quality of the mic vs. the auto gain, noise suppression, and other tweeks? Im having trouble getting it to recognize aliases in home assistant. Im using a logitech USB mic and it recognizes the wake word from across the room but only recognizes the alias when Im 5 ft or less from the mic.
Great Tutorial ! I would love to follow it (if I had enough time; Retirement is still far away ;) ) , I wonder why there are no already built devices like this one to purchase.
So just to make sure i am tracking this correctly we need the Zero W2 ~$25, 2 Mic Pi Hat ~$13 and if we want to avoid soldering the ~$15 hammer kit. So about $55 per speaker. I am tracking this properly. Thanks for all your videos they are excellent and you are helping all of us a ton.
I have been using the Google Home speakers and hubs, but have been waiting for something like this to be able to make a break to Home Assistant. I think that the time to make the change is at hand.
I love the video!! 2 questions: 1. Are you able to connect an external microphone? 2. Can you have it to continue listening until you give a “sleep” command?
I’ve dipped my toes into basically every Home Assistant there is I prefer Apple home, but I bought Home Assistant green, not too long ago going to tinker with that
Dude I LOVE this! Like so many of us out there I am looking for a good solution to replace the Alexa devices in my house. I'm a Sonos household. Is it possible to use this for the microphone and output audio to the sonos speakers? The thought here is to use the high quality speakers I already own. Thoughts here?
Exactly one of the topics I was thinking! I’m going to push Wyoming to its limits (show how to enable those power features) then do a direct comparison.
Hi. Nice one!!!! What's the advantage pros cons of doing HA voice control this way over using Smart Home Junkies ESP32-S3-Box-3 solution and bouncing the voice reply back thru a networked speaker? His solution seems less clunky? Less "parts", more flexible in placement and therefore less likely to break?
Nice Video, although I used latest os for pi zero, had to sudo rpi-update 1440b3e0b52075a9ec244216cddcf56099c28dfa to downgrade to kernel to 6.1.21 to get reseeed to record. Also couldn't do it headless as wlan wouldn't start unless I went in and started it via nmtui manually at 1st install, worked fine after that and fully working now, many thanks for putting this video out.🙂
@FutureProofHomes Can you give me the img file after successfully installing all the steps? I can't install it forever, it keeps giving me an error. thank you.
5:15 Small correction: It is the name of the user you've specified, not the name of the device. The hostname does not affect how this directory is named.
@@FutureProofHomes No worries, good tutorial nevertheless! Just wanted to avoid confusion for linux newbies maybe not finding the directory because of differing host- and/or user names. I really enjoyed the tutorial, keep it up!
This would be perfect if the raspberry pi would be powered by PoE and booting from the network. Also sn Ansible playbook to provision the devices would be awesome :-)
I've heard rumors of a w2 pi that is powered by POE coming in the future. With our luck though they'll make 10 of them and charge 150 a pop. Sucks. I'm curious, can you plug a POE hat and 2-mic pi hat to a Rpi at the same time? That'd be sweet.
I think outputting visual graphics of any type would require some of OpenAI's "DALL·E" magic. That engine/API is still very foreign to me still. I like the creative thinking though!
I am not familiar with pi or linux but looks like something i can do. Currrntly have HA on virtualbox on my main Win11 x64 machine. Some questions. I use home assistant for non cloud based automation since privacy minded, what am i compromising if anything if I access the web(ChatGPT included) for control in this case? Also trying to dechipher why this is a better way as opposed to other local voice command processing out there for home assitant? Thank you.
Hello, thank you very much for this useful video. Can I use another board instead of Re-speaker? I can't find it in my country. do you know any alternatives?
What I'd love is a way to use my old Google Homes. Replace it with this. That way I have a nice cheap system already looking nice. Has a mic and speaker already on Might be a dream but I think it'd be better.
Is there an option to only have microphones in the network, to acieve low power, low impact and set the "intelligence" at one place (docker/vm)? Maybe in a jack box? That should be more efficent as set several Pis up, i suppose?
What you're describing is really close to what HA is talking about in this blog post. Search the page for "Anker PowerConf S330". www.home-assistant.io/blog/2023/10/12/year-of-the-voice-chapter-4-wakewords/#reuse-and-repurpose-different-ways-to-create-a-voice-satellite
another outstanding walkthrough/tutorial!!! question; is there any particular reason you preferred to go with deploying a raspi vs say, an atom echo smart speaker?
You can’t run openwakeword on that much slower/cheaper hardware. And I needed a 3.5mm headphone jack out to integrate in my ceiling and fully replace my in-ceiling Alexas.
Also, the wake word detection on the atom echo is awful (works 2/10 times if the mic isn't pointed at you, and you're extremely close, even then it's like 4/10)
you could also do this on pretty much any old laptop or desktop but need to keep it running all the time, so if its a laptop go for it and put linux then this wyoming thing and follow this tutorial
This is great, I've just ordered all of the hardware. However does it work well without access? I have intermittent internet access so im trying to build fully local home automation around home assistant.
Glad you found the vid helpful! Let us know how your build goes. If you run your entire pipeline locally you will be able to interact with your voice assistant with no internet connection. Watch this video to learn more: ua-cam.com/video/pAKqKTkx5X4/v-deo.html
Nope. To my knowledge TPUs don’t improve the performance of LLMs (inference). They do wonders for video object recognition though. Until someone modifies LLMs they don’t really increase performance of LLMs. VRAM, GPUs & RAM help LLMs.
This is freaking incredible. I'm not so much into giving commands via voice but more getting feedback for things. Like sending TTS messages about time and events that need to happen. Is it possible just to send TTS messages out via the pie and speakers. I'm sure it is possible but I'm not understanding how that would work.
Not yet! Unless you run another service on the raspberry pi that gives it media_player capabilities. This GitHub issue is asking for your exact feature: github.com/rhasspy/wyoming-satellite/issues/40
@@FutureProofHomes That's my one. Hang up with all this local voice stuff. At first talking to Alexa and getting commands was the cool thing to do. But after a while it just got boring and lame. My thing is just having the house automate itself and not having to tell it to do things. My main use case will be to give an announcements when it's time to go to the bus or important announcements about the mailbox or things like that. That's what I'm going to do with local voice. Keep up the good work and I hope when that capability is available you will make an awesome video to go along with your series
Great tutorial! But I cant finish the process. My rpi zero 2w stuck at building wheel for zeroconf. Wheel indicator stops in some point and rpi is not responding. How do I pass that?
This is super cool. Especially your in-ceiling application of it. Is there some way to have the response/output piped through a speaker of your choosing rather than one plugged into the Pi? For example, I trigger the wake word via the Pi Zero and then it responds from a Sonos speaker across the room. Seems like i could be possible with an automation in HA, no?
I fear OpenWakeWord may need more resources than the Pi3 can offer, but try it! I would love if we could direct the TTS response do different speakers though. Check out these links below to track when that capability comes online: github.com/rhasspy/wyoming-satellite/issues/40 github.com/rhasspy/wyoming-satellite/issues/38 github.com/AlexxIT/StreamAssist
I get a very fast and high pitched TTS response using this setup with a small speaker connected via JST. I don’t know which setting affects the speed of the TTS response. Any idea?
Great video!! I am having a problem getting the leds to work .. I’m going to try another hat 1st cause i think thats the issue but question if i wanted to add my own what pins would i use
Wonder if there's a way to connect the Signal messenger add on to the ChatGPT assist. Would be good to receive notifications and converse through signal even away from home
Nice and exciting video. Installed as given in the tutorial and video. As I could not lay my hands on a Respeaker 2 hat , connected a INMP441 which I had used to make a voice assistant with esp32. The wake word does not get recognised. I tried a recording test with the mike and it does record but the volume is low. Could that be the cause for the wake word not being detected? Will a USB microphone work?
Really glad you liked the video! Thank you.. Yes, bad quality audio can definitely be the cause for openwakeword to not respond. I’m assuming you tweaked all the noise suppression, mic volume and gain setting as best you could? Yes, I think a usb mic could possibly work too.
Awesome tutorial, worked like a charm! The only thing not working for me is playing the sound effects when the assistant awakes and is done listening. I added the two lines in the service configuration according to 19:20 and also located the sound files on the pi, it's just not playing though... Any idea why that might be? Thanks a lot and cheers!
The issue i am having with setting up ceiling speakers is actually cutting holes in the ceiling. :) we have a 1 story house and adding all those holes into the attic is not good..Which I had a conditioned attic and then i would not have an issue. Any suggestions of sealing ceiling speakers?
My kids are now old enough to trigger Alexa and ask for ridiculous things. It's driving me nuts, so this is super timely. Edit: Black Cherry is indeed their best flavor. 😂
Create a wake word and pipeline specifically for your kids that responds in a particular voice with a certain personality. They’ll have fun for days. :)
OpenWakeWord has multiple wake words to choose from. You could use "hey jarvis" that responds with a British male voice and a voice assistant pipeline that is capable of controlling the entire home. And then you could create another separate pipeline with the "alexa" wake word that responds with a female voice and limited control to the smart home. When your kids scream "alexa" they can only do certain things. There isn't a way to do automatic voice detection (yet). Just different wake words for different pipelines. Maybe I can do a video on this in the future.
There are quiet a few limitations still. To name a few: 1. No you can't have it stream audio yet (I'm working on it though.. might be my next video actually) 2. The LLM is inconsistent. Sometimes it does exactly what you want, sometimes not. AI is funny like that and eventually it will be solved. 3. It can't yet create alarms, reminders, timers. I think that's coming soon. 4. It requires your entities to be named well. If you feed it light.sdf97asdf9078 it will increase the likelihood weird things happen. There's more, but I'll stop there. Just takes time to sort these things out.
Gotta say, the way you step through these instructions very clearly, with no over the top quirky personality, AND showing multiple issues you can encounter and how to resolve them, is a breath of fresh air in the smart home automation space. Keep up the good work, I'm subscribing to see more!
That means a lot man. Thank you so much for the support!
Completely agree, perfect explanation, detail, and character.
I second that.
Its hard find this many details to work out all kinks of such project. Amazing use of your time sir. Thank you!
Big agree. Subscribed because of it!
Very nice video! I'm the voice guy from Nabu Casa, and the author of the Wyoming protocol. Do you mind if I include this video in the docs?
Absolutely. I'd be honored. Love your work @synesthesiam & pleasure to meet you!
@@FutureProofHomes You as well! I've added links to 2 of your videos in the README :)
the legend.
@@12around1 🤠
Scam
Right to the point. Clear. Concise.
I’m amazed that our most modest devices these days (the Zero) are capable of such useful stuff.
I do believe you are one of the best if not the best UA-cam Tutorial Content Creators I have ever viewed and subscribed to and trust me, my subscription list is very exclusive and normally has to be earned over many videos and seldom is it earned on the first watch. I don't mean to embarrass you but I'll even clarify a few reasons why (for the benefit of other creators that could do worse than take a few tips from here). Verbally, you're easy to understand, clear and concise, I don't have to rewind and replay to hear what you said. You use visuals like highlighting the commands and text that you are talking about. You explain what the command is actually doing which is useful education for people that have blindly used similar commands before but had no idea what they were actually doing AND importantly, these descriptions are just the right length, not to in-depth that lead down a rabbit hole. Finally, you haven't jumped on the band wagon of coming up with a cringy slogan or start every video with an emphasised chant like "let's dive right in" etc. A huge thanks for being more real!!
I predict your channel and subscriptions growing fast and exponentially and deservingly so!
Dang man. Thanks so much for the kind words. That really means a lot to me. (My wife too. She loved your comment.) All the best, sir!
I agree with all this and subscribed on day 1 after building my wyoming board on a pi 4 with USB Mic. I'll add that your github written tutorial is excellent as well. Thanks for taking the time!
Thanks for the tutorial. You don't have to type "y" if want to proceed. If [Y/n] just press Enter for Y, if [y/N] just press Enter for N. Default choice is always a capital letter.
Noted! Nice tip. TY!
We'll have to see how it holds up when things inevitably change in a few months, but for now, this is perhaps the best, easiest true step-by-step tutorial I've ever followed on UA-cam.
I even followed along for things I already know how to do, like exiting nano.
Much love sir! Hope this vid goes far and helps many. Although, in a few years I hope we’re laughing at this tech/video. The future is bright!
Followed these instructions and got an identical config up and running with a Pi Zero 2 W and ReSpeaker 2-Mics Pi HAT. For those in the UK I picked up the ReSpeaker at Mouser. Thanks for the clear step by step instructions, easiest hardware/software project in a while.
Sweet! Glad you got things up and running!!
What speaker / mic did you use, im in UK too.
I had an interesting issue trying get this working but was able to figure it out. I already had followed your ChatGPT instructions and had OpenWakeWord, Piper and Whisper installed. For some reason Home Assistant wouldn't find my new Satellite. I was able to find the instructions to add it manually (using the IP address and port 10700) and then was able to successfully complete the process. I had a set of headphones plugged into the satellite and I'm using the same wake word as your video. Before I installed Open Wake Word locally, it was using what I had already installed in Home Assistant. This is Brilliant! Now to make some changes so that the output of my conversations are redirected to the in-room Sonos speakers!
When you achieve your sonos output please let me know how you did it!
@@Followbrayden there's actually a project Sonos in the module library from HA.
This is awesome. I've been seeking out a whole house local voice assistant solution and this seems to be pretty close to what I desire. Thanks for the inspiration.
Glad it was helpful! Send pics or vid when done!
Hello from Norway!
Thank you sooo much for this. This is brilliant and works like a charm.
Keep up the great work - I really appreciate these awesome videos of yours!
-Best regards Kip
I'm actually moving in the next month and wanted to get away "she who shall not be named." Your videos are going to be my roadmap to a whole new smart home setup. Thank You!!!!!
Love it. Run a small test with the product make sure it suites your needs then scale up to support the entire home. Not trying to be too pushy, but if you need help along your journey I do consulting. ;) Have fun @cliffordraymond9123!
Cool beans! Clear instructions, precise pointers to essential parts and to end a nice demo!
What else can anyone want?
Subscribed!
@TheLMFAOZ (nice username, btw), thanks for the props man! Cheers!
Great tutorial. Ordered parts using your affiliate links - I'm gonna try this out myself. This is much more practical than using an M5Stack Atom Echo, which I've already done. You've got a new subscriber!
I'm actually planning to use the m5stack for a full ai research buddy
why is it more practcal than M5stack?
I've been bothered that out of the box this tutorial (the wyoming-satellite, not your vid) doesn't ship with the ability to do tts notifications. A simple install of mpd solves this nicely - making the communication feel two way rather than just it reply to the wake word.
Good job on the vid! Keep making them.
Thanks for the good video vibes. Yea, you’re right. People definitely want to use this hardware as a media player. FR has been submitted to the Wyoming project here: github.com/rhasspy/wyoming-satellite/issues/40
I’m gonna try installing MPD on the Pi soon so see if I can unlock this feature. Should work.
I came to make the same statement, but you beat me to the clock. It's indeed the only missing piece!
Stay tuned for next vid. ;) ;)
I'm creating a bunch of these around the house, with the eventual plan to repurpose some old Alexa Dots (and some day replacing the rPI with a custom ESP32 board). I control them all remotely via Ansible, which makes life a LOT easier for managing the inventory.
Excellent tutorial. Waiting to see who comes up with the first pre-built ready to go options.
Tempting me? :)
@@FutureProofHomeslol
absolutely
@@FutureProofHomes
Stay tuned... ;) ;)
@@FutureProofHomes is it possible to use Echos with this approach instead of replacing (or supplementing) them with these Wyoming devices? I'm certainly going to get that question from the family.
Okay, I got this setup working on a Raspberry Pi 4, and have it partly configured on two Pi Zero 2 W boards. I already had one ReSpeaker, but I'm waiting on the other two I ordered. Now to take it to the next level and get it working with PulseAudio and Snapcast as a music player.
Love it! Try and setup your first one so you dial it in, then the remaining ones should be easy. Congrats!
@@FutureProofHomes I think I goofed up on something related to PulseAudio on the Pi 4, so I need to figure out what I did wrong and fix it.
Great video, clear & precise instructions, Keep it up your channel will grow just subscribed.
Much love. Thanks man!
I love your videos. Fantastic format and very easy to follow and understand. I am greatly looking forward to more content!
Awesome, thank you!
Wow, what a full step by step explanation. Thanks man!
Glad it was helpful! You’re welcome sir!
@@FutureProofHomes Can you give me the img file after successfully installing all the steps? I can't install it forever, it keeps giving me an error. thank you.
This is the coolest shit ever! Im pretty tech savy and love learning and just started looking at building a home server and home assistant. Im a big user of google and wifi home lights etc but have always dreamed of it being local and always worried about google listening to me lol i love learning and will attempt this tyvm! ❤
yet again ....a clear tutorial...keep up the great work..!
Will do! Thank you!
Bad@$$ bro. I'll be building a home this year, and this will be perfect. I'll watch your linked videos to ensure a stealth install is done properly
Hit me up if you need help. I’ve done it all! :) Good luck sir! It’s awesome when all set up!
Amazing, waited for this one! 2024 will be my year of the voice ;-)
You didn’t say “first”! :)
Thanks for the demo and info. Your instructions are awesome and very detailed. Have a great day
That is really awesome. No NPU needed? Wonder when a touchscreen version become available. Would be awesome.
Great job on the tutorial. Disappointing that this level of configuration is needed for HomeAssistant projects. Yes, it makes it the most flexible home assistant ever..... with an ever-increasing learning curve. Why not just release a SD card image and make a few updates for your settings? Some of us just don't have the time to be HA certified.
I hear you. Some of us just love being punished by hours of software configuration. Time will get use to exactly what you want though.
"why can't software devs just read my mind!"
Nice video, thanks. I have an extra Raspberry Pi 3. Do the steps in this video work with this model or there is something I need to change?
I'm loving your tutorials and everything you are doing. I think you are on the forefront of the future of the smart home and its damn exciting to watch and build alongside you. I've been a HA fan for 7-8 years. I built your FastGPT pipeline last week, pretty fun to use but so far only light usage. I'm curious if you have thoughts on the approximate cost in monthly OpenAI bills if used that FastGPT pipeline heavily to run your smart home, replace the voice assistant and field general queries. I know it's highly dependent on usage obviously, but wondering your thoughts on a ballpark of that.
Is there any way to get a little screen on the satellite to display a home assistant dashboard or text output of the assist conversation? Trying to replace my Google Hub functionality.
Built this and it works like a charm, but you're gonna have to dub out wakewords in your videos from now on as it apparently really likes your voice... 😆
Ha! Okay, noted!
Thanks for the great guide on how to build a Voice Assistant at home. I am currently planning such an installation in my new home and am wondering about the best solutions. What do you think about other Respeaker boards? for example the "ReSpeaker 4-Mic Array for Raspberry Pi" or the "ReSpeaker 6-Mic Circular Array Kit for Raspberry Pi". I want the best possible quality of microphones so that I don't have problems with the wake word and that the board reads commands well. Can these boards be integrated into the HA in an analogous way? Is this a good idea or is it a waste to overpay?
Honestly, the 2-mic Pi hat has served my needs. The 4-mic pi hat is probably better though. Let me know your results! The 6-mic circular array may not work with the drivers I fear, but try it! That all said, I would still recommend the 2-mic hat. It really is comparable to the Alexa/Google based on all my testing. Even works with background noise in a big room if your raise your voice a bit (which I had to do with Alexa as well).
Will background noises be ignored good enough (like running TV)? And can you use any wakeword? And does the software accurately work with German too?
how come nobody had an issue with installing pysliero-vad? I get the error, no matter what: ERROR: Could not find a version that satisfies the requirement onnxruntime=1.10.0 (from pysilero-vad) (from versions: none)
ERROR: No matching distribution found for onnxruntime=1.10.0
I also had it. Did you ever find a fix?
Great tutorial! especially for the visual, makes it easier to understand ---- can you do a video on maybe utilizing the GPIO/12 port for external lighting cue? be nice to have another indicator avenue if we are unable to see the 2-Hat
Awesome work, and easy to follow!
I'm not getting any audio with the extra configurations. Any ideas?
I'd like to add a custom one if possible too...!
Thumbs-up on the video, but I have a small question: Why did you go with "legacy 64-bit" rather than the current 64-bit OS? From what I read, the current lite version does fine on the 2w, but I figured you must have a reason for going with the legacy version.
Bookworm (newest raspi OS) is still rough around the edges and introduces some breaking changes to python env. Just went through safe route to avoid unnecessary headaches.
Another amazing video. I'm in the process of migrating from a heavy Alexa automation to Home Assistant and this solves my Echo device issue. I have a few questions -
1) I have 12 Echo devices and was wondering if I can set up one of the Wyoming Voice Satellite's and then copy that SD for each of the other RPi's and change the name/user info to so that I'm not having to follow these steps on each device?
2) Is there a method to determine the person speaking? I would like to include a response specific to that person
3) In my home automation setup I have Sonos is each room - would it be possible to have each RPi output the audio to the Sonos device in the room where he RPi is via Home Assistant? I agree with your approach on muting the music to allow the voice conversation to happen - that is the behavior I have now with Sonos in the room and the Echo devices. I would love to be able to get the conversation over my Sonos devices in each room.
Keep the great content coming!
1. YES! You can do this. If/when you do, please document the process and share.
2. Give each person their own wake word. I use "hey bob" and you use "hey sally". That's the best way today.
3. My most recent video tries to get rid of Sonos completely. But I also want what you're describing! Yes, this should be possible. Watch my new video and modify the "--snd-command" so that it PulseAudio will use the Airplay protocol to stream to your Sonos. That should be it.
the steps are very clear... though i have now to make it working with french ;) and try to use my local LLM with it... :)
Hello, Just found your channel, this is really cool, keep it up!
Thanks, will do!
Great video, allowed me to get started with this very quickly. At the moment I can't get the LED setting to work, if I enable it then the "awake" is triggered but the unit then just hangs and needs restarting. Have you ever had an issue with the LED? Also, couldn't get "ok nabu" to work but "hey jarvis" is fine - must be my pronunciation!
(Having issues with the ChatGPT pipeline but need to watch you video again to re-set that up - I did it sometime before I got the hardware.)
I appreciate the cool content! Thank you for the inspiration, and keep doing great work! I've just subscribed to your channel :). Now I have to figure out how to get this thing running in my native language :))
Thanks for the sub!
This is awesome! Not having an open mic to the internet is something I have been looking forward to (and so has my wife). About the only thing missing as far as I can tell is the capability to use a custom created wakeword. My programming skills are limited; if it is possible to utilize a custom wakeword please let me know.
It is possible the section where he sets up his wake word to be hello jarvis is where he explains it.
@@pappawolfie The wakewords he is using are not custom createsd; they included in the openwakeword model including Hey Jarvis.
Yes! You can train your own wake word. This should get you started. www.home-assistant.io/voice_control/create_wake_word/
Great tutorial. It's Print Working Directory - pwd. I blame the white claw. As always rtfmp. 8:14
This has been so helpful! Thanks! Where do we go to figure out how to evaluate the quality of the mic vs. the auto gain, noise suppression, and other tweeks? Im having trouble getting it to recognize aliases in home assistant. Im using a logitech USB mic and it recognizes the wake word from across the room but only recognizes the alias when Im 5 ft or less from the mic.
Great Tutorial ! I would love to follow it (if I had enough time; Retirement is still far away ;) ) , I wonder why there are no already built devices like this one to purchase.
I hear you.. I’m sure people are working on it!
@@FutureProofHomes 4 months on still no devices out of the box?
Fabulous! Thanks. Is it possible to use the custom wake word I've created?
So just to make sure i am tracking this correctly we need the Zero W2 ~$25, 2 Mic Pi Hat ~$13 and if we want to avoid soldering the ~$15 hammer kit. So about $55 per speaker. I am tracking this properly. Thanks for all your videos they are excellent and you are helping all of us a ton.
I have been using the Google Home speakers and hubs, but have been waiting for something like this to be able to make a break to Home Assistant. I think that the time to make the change is at hand.
Take a look at the Onju voice project!
Very informative. This is exactly the project I wanted to take on. Subbed 👍 😎
Awesome, thank you! Glad to have you.
I love the video!!
2 questions:
1. Are you able to connect an external microphone?
2. Can you have it to continue listening until you give a “sleep” command?
1. Yes. Can plug USB mic in.
2. Not without some hacking. But nothing is impossible.
I’ve dipped my toes into basically every Home Assistant there is I prefer Apple home, but I bought Home Assistant green, not too long ago going to tinker with that
So you know how deep the rabbit hole goes then? Welcome back to the game! :)
Dude I LOVE this! Like so many of us out there I am looking for a good solution to replace the Alexa devices in my house. I'm a Sonos household. Is it possible to use this for the microphone and output audio to the sonos speakers? The thought here is to use the high quality speakers I already own. Thoughts here?
Great tutorial, everybody wants Jarvis in his home ^_^ Subscribed to see more of your content,...keep 'm coming!
Much love. Thanks man!
Really appreciate your research and the well made videos, Thank's a lot!
Thank you for that!
Amazing video!! After some use, It would be greate if you can provide some pros and cons in comparison with Alexa and google devices. Could you? 😊
Exactly one of the topics I was thinking! I’m going to push Wyoming to its limits (show how to enable those power features) then do a direct comparison.
This so well documented!. Thanks for this video❤
ok man i gotta know where you found the rj45 to power and 3.5mm jack adapter? ive been looking forever and cant find anything like this to do this,
also how are you using a mic with this?
Hi. Nice one!!!! What's the advantage pros cons of doing HA voice control this way over using Smart Home Junkies ESP32-S3-Box-3 solution and bouncing the voice reply back thru a networked speaker? His solution seems less clunky? Less "parts", more flexible in placement and therefore less likely to break?
Great! Would be nice to have external mics in different rooms aswell
Nice Video, although I used latest os for pi zero, had to sudo rpi-update 1440b3e0b52075a9ec244216cddcf56099c28dfa to downgrade to kernel to 6.1.21 to get reseeed to record. Also couldn't do it headless as wlan wouldn't start unless I went in and started it via nmtui manually at 1st install, worked fine after that and fully working now, many thanks for putting this video out.🙂
@FutureProofHomes
Can you give me the img file after successfully installing all the steps? I can't install it forever, it keeps giving me an error. thank you.
5:15 Small correction: It is the name of the user you've specified, not the name of the device. The hostname does not affect how this directory is named.
Yea, I saw that too! Sorry guys. I’m freestyling sometimes during these recordings and we’re all human. I’ll keep an eye out next time!
@@FutureProofHomes No worries, good tutorial nevertheless! Just wanted to avoid confusion for linux newbies maybe not finding the directory because of differing host- and/or user names.
I really enjoyed the tutorial, keep it up!
What speakers / microphones are people using?
This would be perfect if the raspberry pi would be powered by PoE and booting from the network. Also sn Ansible playbook to provision the devices would be awesome :-)
You can get a poe hat for the raspberry pi but I don't think it works on the zero.
I've heard rumors of a w2 pi that is powered by POE coming in the future. With our luck though they'll make 10 of them and charge 150 a pop. Sucks. I'm curious, can you plug a POE hat and 2-mic pi hat to a Rpi at the same time? That'd be sweet.
Another excellent vid! Thank you! I can't help but wonder if ChatGPT could "visualilze" your house floorplan from sensor data...
I think outputting visual graphics of any type would require some of OpenAI's "DALL·E" magic. That engine/API is still very foreign to me still. I like the creative thinking though!
awesome great video it was super helpful :) i'll be going back and watching some of your other videos on this voice assistant for home assistant.
Wonderful. As any questions as you go backward in time on watching the videos. Here to help!
Very well made intro
I have quite a lot spare RPI 3. Can't wait to try this.
Keep us posted on how it goes!
Can you recommend a good tutorial for Home Assistant? Feels like I should set that up, first.
Got this working and its working flawless.
Is there a way to use the button on the ReSpeaker 2Mic HAT in home assistant?
I am not familiar with pi or linux but looks like something i can do. Currrntly have HA on virtualbox on my main Win11 x64 machine. Some questions. I use home assistant for non cloud based automation since privacy minded, what am i compromising if anything if I access the web(ChatGPT included) for control in this case? Also trying to dechipher why this is a better way as opposed to other local voice command processing out there for home assitant? Thank you.
Hello, thank you very much for this useful video.
Can I use another board instead of Re-speaker?
I can't find it in my country. do you know any alternatives?
Nice walkthrough, thanks for your work
My pleasure!
What I'd love is a way to use my old Google Homes. Replace it with this. That way I have a nice cheap system already looking nice. Has a mic and speaker already on
Might be a dream but I think it'd be better.
Google search “Onju Google Home”. That project is heading in the direction you want. Requires hardware skillz tho. Good luck!
Is there an option to only have microphones in the network, to acieve low power, low impact and set the "intelligence" at one place (docker/vm)? Maybe in a jack box?
That should be more efficent as set several Pis up, i suppose?
What you're describing is really close to what HA is talking about in this blog post. Search the page for "Anker PowerConf S330".
www.home-assistant.io/blog/2023/10/12/year-of-the-voice-chapter-4-wakewords/#reuse-and-repurpose-different-ways-to-create-a-voice-satellite
@@FutureProofHomes thanks i will looking into that. :)
another outstanding walkthrough/tutorial!!!
question; is there any particular reason you preferred to go with deploying a raspi vs say, an atom echo smart speaker?
You can’t run openwakeword on that much slower/cheaper hardware. And I needed a 3.5mm headphone jack out to integrate in my ceiling and fully replace my in-ceiling Alexas.
Also, the wake word detection on the atom echo is awful (works 2/10 times if the mic isn't pointed at you, and you're extremely close, even then it's like 4/10)
Is there a way to weak the speed of the output voice? The sound is spoken extremele slow.
There is. I can remember the exact parameter to add, but it’s possible! Dig through the documentation.
you could also do this on pretty much any old laptop or desktop but need to keep it running all the time, so if its a laptop go for it and put linux then this wyoming thing and follow this tutorial
True!
Frickin awesome. Going to see if my RPI 4B can run this alongside HAss. Work to do....
Get after it! Glad you liked the vid man. Thank you.
This is great, I've just ordered all of the hardware. However does it work well without access? I have intermittent internet access so im trying to build fully local home automation around home assistant.
Glad you found the vid helpful! Let us know how your build goes.
If you run your entire pipeline locally you will be able to interact with your voice assistant with no internet connection. Watch this video to learn more: ua-cam.com/video/pAKqKTkx5X4/v-deo.html
Thanks a lot for the video!
One question, would you consider coral TPU to improve performance?
Nope. To my knowledge TPUs don’t improve the performance of LLMs (inference). They do wonders for video object recognition though. Until someone modifies LLMs they don’t really increase performance of LLMs. VRAM, GPUs & RAM help LLMs.
Cool, thanks a lot!
Cool, thanks a lot!
Great video! Subscribed!
Awesome, thank you!
This is freaking incredible. I'm not so much into giving commands via voice but more getting feedback for things. Like sending TTS messages about time and events that need to happen. Is it possible just to send TTS messages out via the pie and speakers. I'm sure it is possible but I'm not understanding how that would work.
Not yet! Unless you run another service on the raspberry pi that gives it media_player capabilities. This GitHub issue is asking for your exact feature: github.com/rhasspy/wyoming-satellite/issues/40
When this feature comes along it COULD replace Sonos. ;) ;)
@@FutureProofHomes That's my one. Hang up with all this local voice stuff. At first talking to Alexa and getting commands was the cool thing to do. But after a while it just got boring and lame. My thing is just having the house automate itself and not having to tell it to do things. My main use case will be to give an announcements when it's time to go to the bus or important announcements about the mailbox or things like that. That's what I'm going to do with local voice. Keep up the good work and I hope when that capability is available you will make an awesome video to go along with your series
Ha! Thanks for requesting that feature. I will certainly make a vid about replacing Sonos when it’s built into my voice assistants. Game changer!!
@@rasbe6863 Totally agree. 4 months on do we have this feature yet?
Great tutorial! But I cant finish the process. My rpi zero 2w stuck at building wheel for zeroconf. Wheel indicator stops in some point and rpi is not responding. How do I pass that?
This is super cool. Especially your in-ceiling application of it. Is there some way to have the response/output piped through a speaker of your choosing rather than one plugged into the Pi?
For example, I trigger the wake word via the Pi Zero and then it responds from a Sonos speaker across the room. Seems like i could be possible with an automation in HA, no?
Also, I'm assuming this could run on a Pi3, too, no? I have a spare one kicking about in a drawer...
I fear OpenWakeWord may need more resources than the Pi3 can offer, but try it!
I would love if we could direct the TTS response do different speakers though. Check out these links below to track when that capability comes online:
github.com/rhasspy/wyoming-satellite/issues/40
github.com/rhasspy/wyoming-satellite/issues/38
github.com/AlexxIT/StreamAssist
I get a very fast and high pitched TTS response using this setup with a small speaker connected via JST. I don’t know which setting affects the speed of the TTS response. Any idea?
Great video!! I am having a problem getting the leds to work .. I’m going to try another hat 1st cause i think thats the issue but question if i wanted to add my own what pins would i use
I had the same issue.
@@danielkedinger it turned out to be bad top hat leds for me .. i swapped it and they worked
Wonder if there's a way to connect the Signal messenger add on to the ChatGPT assist. Would be good to receive notifications and converse through signal even away from home
Nice video. I want to try it. Do you know what version of the os you are using? I’m having a helluva time getting the respeaker drivers installed.
Nice and exciting video. Installed as given in the tutorial and video. As I could not lay my hands on a Respeaker 2 hat , connected a INMP441 which I had used to make a voice assistant with esp32. The wake word does not get recognised. I tried a recording test with the mike and it does record but the volume is low. Could that be the cause for the wake word not being detected? Will a USB microphone work?
Really glad you liked the video! Thank you.. Yes, bad quality audio can definitely be the cause for openwakeword to not respond. I’m assuming you tweaked all the noise suppression, mic volume and gain setting as best you could? Yes, I think a usb mic could possibly work too.
Awesome tutorial, worked like a charm! The only thing not working for me is playing the sound effects when the assistant awakes and is done listening. I added the two lines in the service configuration according to 19:20 and also located the sound files on the pi, it's just not playing though... Any idea why that might be? Thanks a lot and cheers!
The issue i am having with setting up ceiling speakers is actually cutting holes in the ceiling. :) we have a 1 story house and adding all those holes into the attic is not good..Which I had a conditioned attic and then i would not have an issue. Any suggestions of sealing ceiling speakers?
how can you control the leds they are bright everything else perfect thank you
wow your skill is amazing Sir
Why thank you @wiwingmargahayu6831!
How are you managing with background noise and/or music from your satalite's speaker?
My kids are now old enough to trigger Alexa and ask for ridiculous things. It's driving me nuts, so this is super timely.
Edit: Black Cherry is indeed their best flavor. 😂
Create a wake word and pipeline specifically for your kids that responds in a particular voice with a certain personality. They’ll have fun for days. :)
@@FutureProofHomeshow do you set this? As in for the voice assistant to recognize different voices.
OpenWakeWord has multiple wake words to choose from. You could use "hey jarvis" that responds with a British male voice and a voice assistant pipeline that is capable of controlling the entire home. And then you could create another separate pipeline with the "alexa" wake word that responds with a female voice and limited control to the smart home. When your kids scream "alexa" they can only do certain things.
There isn't a way to do automatic voice detection (yet). Just different wake words for different pipelines. Maybe I can do a video on this in the future.
Hi there, another great video, very much impressed 😎
So, what are the limitations?
Can you still play music from streaming providers via Jarvis?
There are quiet a few limitations still. To name a few:
1. No you can't have it stream audio yet (I'm working on it though.. might be my next video actually)
2. The LLM is inconsistent. Sometimes it does exactly what you want, sometimes not. AI is funny like that and eventually it will be solved.
3. It can't yet create alarms, reminders, timers. I think that's coming soon.
4. It requires your entities to be named well. If you feed it light.sdf97asdf9078 it will increase the likelihood weird things happen.
There's more, but I'll stop there. Just takes time to sort these things out.
Hi has anyone had problems with reboots or just stop working after a day or so?
Is it possible to clone the micro sd card to another pi? So you can have two satellites