Protect yourself, your family and your business from scams: bitdefend.me/SCNetworkChuck I’m replacing Amazon Alexa with my own, completely local AI voice assistant!! The amazing part is that it cleanly integrates with my home automation system, Home Assistant. Also, it’s using local LLMs like llama3.2 with Ollama. I think it’s finally time for you to replace your voice assistant with something local. 📜📜📜COMMANDS: academy.networkchuck.com/blog/local-ai-voice-assistant 🔥🔥Join the NetworkChuck Academy!: ntck.co/NCAcademy **Sponsored by Bitdefender
Hey Chuck, I have an issue with people coming by my house for made up reasons and then somehow using their cellphones and somehow getting access to either my wifi or they are setting up some sort of wifi tunnel. I think my live in girlfriend is somehow helping them. Now I have very limited knowledge of how WIFI and the internet works, all I know is far too many people have information about me that they should not have.
Creating a custom voice is a multi-step process. I previously cloned a female English voice and adapted it for Brazilian Portuguese and Italian, generating .onnx models for each. My method involved creating a dataset of sentences using a text-to-speech (TTS) model already trained on my target voice. This dataset was then used to train my custom voices using substantial computing resources. Specifically, I believe I used a fork of Coqui TTS to generate the sentences, running it on a LightningAI server equipped with four NVIDIA L40S GPUs. The .onnx models were then created following Piper's guidelines. More training leads to better voice quality, but requires a significant time investment.
Finally, you have now replicated the journey i take with homelabbing... Pure pain and suffering on every single step of the journey just to get a sort of working result and most of the time doesn't even work.
Wait, you are taking linus advices seriously? Bruh you have a problem. Anyone with an above average IT knowledge know his channel wnd content are for entertainment, and nobody should try to use or implement anything that comes from that channel, rip.
@@theangelofspace155 well, just a counter point, I decided to build my home lab because of a Linus video. Then went to HH, Techno Tim, Raid Owl, Wendel both Jeffs, and some very good HA channels. I know better than to look to Linus for how to. But his channel is pretty good for inspiration..... and of course entertainment as you said.
Hi Chuck. Thanks for the end-to-end on using Voice Assistance. BTW, when you activated the wake-up word "Hey MyCroft", then demonstrate using "OK MyCroft", it's not surprising when it doesn't work 😉😊
Dude. I have followed multiple guides. Banging my head. I noticed you didn't install a wakeword service on the satellite, which I was doing. I just omitted all that stuff and it worked flawlessly! Thank you!
If you have dedicated hardware for your "satellite" it is also possible to run the wake word detection locally. Then you have less network traffic and less lag in the voice assistant.
Місяць тому+21
Mycroft AI business went down but the software got forked and its actively maintained by the Open Voice OS and Neon communities. I'm running my Open Voice OS on my laptop (but it can run on Raspberry Pi as well) with a local LLM! Thanks for showing people that there is something else than the big Alexa and others!
OMG. Paulus Schoutsen and Mike Hansen have been talking about this since last year, the "Year of the Voice" You have inspired me to do it! Thanks, Chuck!
YESSS more home assistant videos please 🙌 I won't have a chance to build the things I want until the summer (2025) so hopefully a couple more home assistant videos drop between now and then so I can re-watch them!!
Great video! One important thing to note: If you want another model instead of llama just make sure it supports Tool calling, it's tagged with "tools" on Ollama models list.
I completely replaced Alexa with HA Assist 3 months ago. At that point it was ready for maybe 30 percents, but i wanted to give myself a reason to speed up the development. Now it does everything we wanted from Alexa and more.
This video and your previous one are absolutely wonderful. I love the fact that everything runs locally without an external 'cloud'. Looking forward to the upcoming videos.
You are not old enough for this reference, but I plan to train mine to say "compliance" or "unable to comply" as the reply for executing tasks. Thank you for these videos - I'm right behind you! Have been looking for this for years now.
No kidding. I've gone from expecting to get things done in minutes in windows to being happy I blew a whole weekend getting something running in Linux. These people really love their dos console window. It's like they skipped the GUI revolution of Windows and just made a 64 bit multi-threaded version of DOS.
Bro, I had to pause the video at 5min to write this immediately. I have been research in my own spare time for the past 3 months on HA and intergrating Voice, AI and control with absolutely no luck. You seem to have very eloquently explained exactly what I need to know (without watching the whole video yet). seems I can pull the trigger on buying stuff for this project and get to it. I'll watch the rest of this, then your HA tutorial and it looks like you've got a sub.
Inspiring video! With home assistant I bet you can up the spouse approval but having your voice assistant say something like "let me think about that" and then process the question.This would pad the wait time for the AI to process the request.
Awesome video. In the process of following along now. Just wanted to say how grateful I am that you removed the slurp sound from your sip of coffee. It's wonderful to see the rising level of misophonia awareness.
Man! Last month I started building my own APIs using node red and LLama. Now I'm there and controlling many things sending backend calls to my api connected to my Arduino board, That makes my workflow pretty much easier! Thanks!
A lot of work went into this video. Well done. This is a real gem. I have the same pi hat and gave up trying to get it working. Think I threw it against a wall. Going to give it another shot if it still works. Thank you!!
Hey, one thing I did was the use of Assist and GPT in the same command, Assist is waaaay faster to respond to obvious commmand like "turn on the light", but it cannot do AI, so I did a nodered script that use assist and if it fails goes to GPT, it works quite well to be fair
Dude, right on, right on. Use to watch your videos all the time, and now I'm stuck again trying to learn some voice recognition stuff the past few weeks, and you just came in to save the day! Thank you!
22:32 Can confirm that ollama binds its port to 127.0.0.1 on debian 12 as well and you need to configure it to bind to 0.0.0.0 or any specific network interface as well if you want to access it from other machines.
That is about the coolest thing I've seen in home automation, ever. Don't think I'll do it though. I can copy what you did, and I'm sure it will work. Until some of that software changes. Then, I'll have no idea how to fix it on my own. I'd love to see links to resources showing how you learned to do all of this stuff.
33:35 It might be a good idea adding something like "be concize in the response" to the LLM system prompt so that you do not get crazy long responces read out 🤣
This is amazing you give great step by step description and easy to follow along guide... I am so surprised by the things home assistant is capable of..Please keep the home assistant videos coming
If you want to build your own hockey puck/snowball, someone already did that, too. I don't recall what all it entails, as I found this article like 2 months ago, but it uses an ESP32 instead of a Raspberry Pi to connect to Home Assistant. Hope this helps someone. Great work, Chuck!
Thank you for muting the drinking sounds, I braced for impact and was hit with delightful silence. It was noticed and appreciated! My Misophonia is evil.
Lack of a ready-to-go voice assistant, and hardware for it, is the reason I've not been able to commit to Home Assistant. Rhasspy may almost have the software, but the cost to build dedicated devices to replace the 8 Google Home receivers around my house is somewhat impractical. However, this video is the best solution I've seen so far!
It occurs to me that you don't need to have a raspberry pi running each individual reciever. Simply set up wireless microphones in place of the various google homes and have them all connected to a central pi. You're going to spend a bit more than the price of a google home, but a fraction of the cost of 8 of them. Considering the power consumption is so low you could have the mics battery powered and have them all over your house if you wanted. It also strikes me that you don't really need them in every room since you can just use your phone 99% of the time. You really only need them in the main living space(s).
Chuck and gang, I am banging my head against the wall here. I have whisper and piper running on a windows computer, followed your directions but when I go back to add the service, I get connection failed. I have tried everything I know including adding the ports to the advanced firewall settings EVEN though they are turned off. What the heck am I missing here. Love the video and thank you for all you do. This is going to step my entire house up.
did you do definitely do the localhost step to 0.0.0.0?? I couldn't get ollama to connect without this. no amount of ports etc made any difference for me.
That is not a Gen 1 Alexa. My daughter purchased me a Gen 1 Alexa for Christmas the year they come out. I still have it running in my bedroom. We moved across country last year. The unit is paired, but it is not possible to re-pair it. Therefore my local weather and time on the device are wrong. Thank you for this video Chuck. My wife is handicapped and we use Alexa's all over the place and I wanted to replace them with my own setup. For bonus points, we listen to music with an MPD client/server setup. If there was someway to voice control our local mpd collection that would be awesome!
Looking forward for someone who gonna make some PCBs to replace some echo devices guts or, maybe, some firmware replacement. The lights, mics and speakers are better, form factor too. All have built-in wifi and some have the screens/displays. Just the cloud/IA stuff that sucks. Taking back the hardware we already paid for would be amazing
This. My dream smart speaker would be some HA-enabled board like esp32 or pico inside a HomePod mini, because those speakers are mind-blowing, they have a pretty clean, full sound with incredibly fat low frequencies considering the small form factor. Maybe even go full Frankenstein and swap the mic for the one that's in Google's Nest mini, this little guy impresses me with how far it can catch my voice from. It would definitely not be cheap tho and quite hard to do, if at all possible. But hey, one can dream, right?
I actually worked with one of the original Mycroft creators earlier in my career. First time I ever saw a 3D printer in real life was at their Alpha launch/Christmas party. It was a fun project to follow along with. That said, I'm in the process of doing exactly what you did in this video. I'm running Whisper, Piper, and Ollama on my TrueNAS instance, and I'm running HA baremetal on an old i3 Gigabyte Brix.
I have tried doing this and it was so slow and decided not to use it. I had no idea you could off load whisper and piper - GENIUS!! And the Ollama for an LLM - I need to read more. Please more of these if you can!! Thank you!! One question - instead of a GPU could a TPU work like google Corel Accelerator? But thank you - more please!!
@@MrThesoulripper13 @trekbear69 current Hailo/Coral TPUs are sadly ONLY good for AI Vision, even just for object detection Frigate BUT NOT for Training, for LLM You need nVidia GPUs, however good news Ollama can now supports newer Intel GPU Iris as well, whisper NOT yet
Awesome video Chuck.. I'm actually in the process of adding features to my HA installation, so I'm really loving these ideas. Especially when it involves replacing Alexa with something that runs locally.
If you eventually decide to containerize Home Assistant and move it to Docker or better yet Kubernetes (would probably make for a good video), just be aware that Add-ons are more limited. You can still install them, but they need to be installed into their own containers and cannot be installed through the HA Add-ons screen, so there's a few more hoops to jump through.
Why would you do that?? Install Proxmox on the machine and make a HA VM, if you need GPU acceleration you can easily just passthrough it... DO NOT use docker for HA...
Can I install all this whisper stuff in a docker? I have Home assistant running on my synology, but I want to replace the alexas with raspberry pi's. Or, should I jsut run it all off the raspberry pi?
@@TuckerDansieFilms Because its more of a PITB for doing add-ons in docker Though I do run it in docker myself - but I also don't really use add-ons...I think I use one but then I might also have just given up too
I am having some trouble if anyone can help. I'm at about 26:03 (ish) and I can run the docker command fine and the whisper service is running> I get the INFO:__main__:Ready but when I go to put in my IP and port 10300 it cannot connect. I see my WSL is running a different IP than my PC but I don't know enough about all this to understand why Home Assistant can't find my whisper instance running on port 10300 when it's active and running. Does anyone have any suggestions?
GPU acceleration with docker whisper is pretty cool. Though Llama3.1 70B eats up so much of my VRAM that it does not fit. It enabled you to run the medium-int8 model which is a lot more accurate and fast with GPU. I'm currently trying to get AMD GPU acceleration to work with my Proxmox with AMD GPU on a Ryzen 7 4750. This is the coolest project ever. My voice assistant mocks and insults me relentlessly. I'm so happy!
Very cool Chuck, I am just now getting started with your academy (the free version) I wish I found your videos sooner nonetheless it is a great time to be alive. I am an electronics technician retired USN my interests now are in Ai and learning IT to become a network engineer or at least an aspiring one. I will be running my own corporation as well as it is just now starting up. A lot of what I have learned in my time (just being alive) is I dont need to know everything to be a leader in any field just a good headhunter. To do this I must know more than the basics and for this type of business my mind is like a dry sponge. Your videos are inspiring and easy to follow. I have come up with my own ideas about practical use of AI and now I see many people have already been working on this. I am encouraged as most people still dont know this is the perfect time to be in IT Networking and AI development. Thanks to you my journey is becoming much smoother. Thanks and I will be subscribing and following you on youtube.
I'm in the Apple ecosystem and have been using HA for years. Their homebridge is incredible and actually allows me to just use siri to control everything I expose to it. Devices, scenes etc.
I love what Nabu Casa are doing with Home Assistant. I actually really loved the telephone gimmick they have where you can use an IP Phone to talk to the voice assistant. I know most people wouldn't use it, but something about picking up a phone to call home assistant was kind of fun.
thats interesting. if u can help with some links or something it will be helpful. also i have a question if u know Im new on this world :) I have a raspberry pi 5 / 8gb Ram and installed the HA on an sd card 512gb U2 class 10. When i installed this addons all the system crashed and a message appeared that is out of network. I do this again and again and the same msg appears do u have any idea what its going on ? i know u cannot help me with this info but am asking for an opinion if anybody knows.
@@martinospapantoniou4491 UA-cam is very good at deleting comments that have links etc. I hate doing this, but it's the best I have. Do a quick search for voice and home assistant. :) Wish I could give more info.....
Chuck you can make aliases of your devices in home assistant and call the what ever and expose them to your pipeline. It's kind of a pain in the butt and a long process, but at least you could call your lamp whatever you wanted to when you talked to your voice Assistant
What has the YT algorithm done to you? The over excited gestures and speech that you see in ipad kids. Fight it don't let the numbers change you, its not to late.
Sadly ollama (AI in general) needs lots of power to work Alexa like. For someone like you, who has a power server to host that, thats nice, even a bot on telegram to have your own AI. But for the common guy with only a raspberry thats not doable (yet) Glad to see it working fine (as long as you can provide resources to it)
One of the commands we use with Alexa is to start timers. Sometimes we have multiple timers running at the same time. Can you show an example of starting a 2 minute timer, along with a spoken phrase of "your timer is done" (or "your egg timer is done" if you've given the timer a name of egg). We tell Alexa to start a timer by saying "alexa, start a 5 minute oven timer". It is fine to have Terry announce completion once and turn off, as opposed to alexa that continues to alert until you tell it to cancel. Also need a way to cancel a timer in progress.
"I don't care how much coffee it takes; I'm going to do it!" LOL This is the second video of yours that I have watched, and I now understand why you have over 4 million subscribers. Pretty amazing stuff. You really know your coffee. Keep up the great work; you won another subscriber. I can't wait to set up the voice assistant on my HA system.
I like how he says that he does not want some other party to have his data, but 80 percent of his other videos are building something that can run flawlessly on a local machine, but he makes it online on some random VM somewhere in the world :)
I'm speechless. I've discovered your channel because of this video and I can't even follow your speed just watching you writing code XD. I'll try to follow your steps just to achieve the dream of having my own "BT-7274" as an assistant. Thank you so much for shedding light on this topic and I'll be waiting for your "Voice change" tutorial. I've already extracted the BT voice lines from Titanfall 2 but I wasn't sure of some steps explained in the documentation. Hopefully, you will know how to do it. And again, thank you so much for this content. New sub!!!
Man, I've just bought HA Yellow , and got to look into this vid - awesome. Gonna bookmark it and implement tomorrow. Planning to run Ollama thing on my Z420 server, in a separate VM.
Something you might try to add to these videos using docker is using compose or even portainer. Makes managing docker in general easier, especially if you are doing multiple containers as it helps visualize things a lot better.
Shouldn't it also be possible to accelerate everything with the Raspberry Pi 5 and the AI HAT+ if you don't have an AI workstation? That would be interesting to know.
34:10 Would have been super useful to have a side-by-side or timed comparasing of the 3 setups! all on raspberry vs offload to laptop vs offload to terry. That way you get an idea on what is worse case, average and best case scenario! A followup would be nice.
My brain just melted when I just noticed I could possibly re use the speakers I don’t use on my 3 seater excalibur / melendez couch. One for the pi / mic, one for the speaker
Totally awesome. My home assistant is already running on docker. Let's hope that my good old QNAP is able to run ollama. But since HA addons are also docker containers inside HAOS, you just need to run your HA on a better HW like a mini pc
Love your work Chuck. You mentioned that you don’t use frigate. I would love to see it though with home assistant and local AI, when object, person detected etc it send an alert with AI generated description of what it sees
This video is awesome. I spent way too much time trying to offload all of the AI stuff to my other server only to get lost in docker stuff. Thanks for putting this out there. Also, getting a custom voice setup is nearly as fun as removing a foley catheter while its inflated. I was ultimately successful, but the voice sucked and the juice really wasnt worth the squeeze. My hope of getting into an argument with Dagoth Ur, the god, have been postponed.
I thought you were going to have some servios closing the doors and raising and lowering the toilet lids In your home assistant video ..but nice vid anyhow lol Love the video!
This is absolutely wow! Would you consider doing a special "upgrade" of Terry incorporating an n8n AI pipeline with more datasources, local folders/databases and things using Home Assistant? Is this even possible? Would probably be like putting "something" between HA's conversational agent and Ollama, right?
1:21 That's the first widely available version of the dot, but it's the 2nd model overall. The very first was only available to purchase using a voice command with the original pringles can sized echo. It was all metal and only had the wake and mute buttons because the volume was controlled by twisting the top. They were pretty awesome, but they sounded like crap and they were inordinately expensive compared to what they charge for a dot nowadays. I paid $80 IIRC.
Super excited if he gets the Terry voice going. I've been wanting to train mine to sound like a transylvanian lab assistant. I named him Igor. Id like to get as close to Marty Feldman's character from Young Frankenstein as possible.
Good luck with changing the AI voice. I was watching the STNG episode, "Skin of Evil," and thought the voice of Armus would be funny for an voice assistant.
Protect yourself, your family and your business from scams: bitdefend.me/SCNetworkChuck
I’m replacing Amazon Alexa with my own, completely local AI voice assistant!! The amazing part is that it cleanly integrates with my home automation system, Home Assistant. Also, it’s using local LLMs like llama3.2 with Ollama. I think it’s finally time for you to replace your voice assistant with something local.
📜📜📜COMMANDS: academy.networkchuck.com/blog/local-ai-voice-assistant
🔥🔥Join the NetworkChuck Academy!: ntck.co/NCAcademy
**Sponsored by Bitdefender
Hey Chuck, will a zimaboard work for this?
Hey Chuck, I have an issue with people coming by my house for made up reasons and then somehow using their cellphones and somehow getting access to either my wifi or they are setting up some sort of wifi tunnel. I think my live in girlfriend is somehow helping them. Now I have very limited knowledge of how WIFI and the internet works, all I know is far too many people have information about me that they should not have.
Creating a custom voice is a multi-step process. I previously cloned a female English voice and adapted it for Brazilian Portuguese and Italian, generating .onnx models for each. My method involved creating a dataset of sentences using a text-to-speech (TTS) model already trained on my target voice. This dataset was then used to train my custom voices using substantial computing resources. Specifically, I believe I used a fork of Coqui TTS to generate the sentences, running it on a LightningAI server equipped with four NVIDIA L40S GPUs. The .onnx models were then created following Piper's guidelines.
More training leads to better voice quality, but requires a significant time investment.
Finally, you have now replicated the journey i take with homelabbing... Pure pain and suffering on every single step of the journey just to get a sort of working result and most of the time doesn't even work.
Can you help me please...🙏
My Gmail account was hacked 😭😭
Between you and Linus' home automation series, I am living the dream when it comes to video tutorials on how to create a locally ran smart home.
Wait, you are taking linus advices seriously? Bruh you have a problem. Anyone with an above average IT knowledge know his channel wnd content are for entertainment, and nobody should try to use or implement anything that comes from that channel, rip.
@@theangelofspace155 can you recommend some other channels please
@@theangelofspace155 well, just a counter point, I decided to build my home lab because of a Linus video. Then went to HH, Techno Tim, Raid Owl, Wendel both Jeffs, and some very good HA channels. I know better than to look to Linus for how to. But his channel is pretty good for inspiration..... and of course entertainment as you said.
@@theangelofspace155bruh you gotta chill
LTT is a decent starting point they aren't as detailed but i use them for overviews
There’s something weirdly refreshing about watching videos like these where you understand absolutely nothing going on but still fascinated by it
Same..
same =)
Yeeep
Same:)
bro I understand what hes saying and its still fucking awesome. just keep listening and learning :)
Hi Chuck. Thanks for the end-to-end on using Voice Assistance.
BTW, when you activated the wake-up word "Hey MyCroft", then demonstrate using "OK MyCroft", it's not surprising when it doesn't work 😉😊
Dude. I have followed multiple guides. Banging my head. I noticed you didn't install a wakeword service on the satellite, which I was doing. I just omitted all that stuff and it worked flawlessly! Thank you!
hi, what you do, can you tell me please?
If you have dedicated hardware for your "satellite" it is also possible to run the wake word detection locally. Then you have less network traffic and less lag in the voice assistant.
Mycroft AI business went down but the software got forked and its actively maintained by the Open Voice OS and Neon communities. I'm running my Open Voice OS on my laptop (but it can run on Raspberry Pi as well) with a local LLM! Thanks for showing people that there is something else than the big Alexa and others!
OMG. Paulus Schoutsen and Mike Hansen have been talking about this since last year, the "Year of the Voice" You have inspired me to do it! Thanks, Chuck!
YESSS more home assistant videos please 🙌
I won't have a chance to build the things I want until the summer (2025) so hopefully a couple more home assistant videos drop between now and then so I can re-watch them!!
Cool, I'm currently buying some sensors right now
By then all this tech will be obsolete.
We'll have 1usd self replicating nano robots conquering the world, lol.
@@ronilevarez901 this us so inaccurate!
The robots will be kidnapping humans and selling them
Du bist mega! Deine Energie ist toll anzusehen. Und dabei auch noch was zu lernen. Grüße aus Deutschland! 🇩🇪
Great video! One important thing to note: If you want another model instead of llama just make sure it supports Tool calling, it's tagged with "tools" on Ollama models list.
I completely replaced Alexa with HA Assist 3 months ago. At that point it was ready for maybe 30 percents, but i wanted to give myself a reason to speed up the development. Now it does everything we wanted from Alexa and more.
This video and your previous one are absolutely wonderful. I love the fact that everything runs locally without an external 'cloud'. Looking forward to the upcoming videos.
You are not old enough for this reference, but I plan to train mine to say "compliance" or "unable to comply" as the reply for executing tasks.
Thank you for these videos - I'm right behind you! Have been looking for this for years now.
Seeing Network Chuck breakdown before tech challanges really helps in taking the edge off my own tech oriented failures :D
No kidding. I've gone from expecting to get things done in minutes in windows to being happy I blew a whole weekend getting something running in Linux. These people really love their dos console window. It's like they skipped the GUI revolution of Windows and just made a 64 bit multi-threaded version of DOS.
I WAS JUST TALKING ABOUT THIS YESTERDAY!!!! Crazy you posted this.
Same!
Same
Same
SAME HOW IS THAT POSSIBLE
*Coincidence?*
Same
Lol I love how chuck reacted when he heard the revision hahahhaha That's the most genuine reaction I have seen in youtube this year
Bro, I had to pause the video at 5min to write this immediately. I have been research in my own spare time for the past 3 months on HA and intergrating Voice, AI and control with absolutely no luck. You seem to have very eloquently explained exactly what I need to know (without watching the whole video yet). seems I can pull the trigger on buying stuff for this project and get to it. I'll watch the rest of this, then your HA tutorial and it looks like you've got a sub.
Inspiring video! With home assistant I bet you can up the spouse approval but having your voice assistant say something like "let me think about that" and then process the question.This would pad the wait time for the AI to process the request.
Awesome tutorial. I'm getting home assist going and trying to use this through the Google home speakers I already have. Not sure if this is possible
this is really cool, not fully there yet, but a lot better at answering questions than alexa
Awesome video. In the process of following along now. Just wanted to say how grateful I am that you removed the slurp sound from your sip of coffee. It's wonderful to see the rising level of misophonia awareness.
Thank you for this amazing Video ❤❤
Man! Last month I started building my own APIs using node red and LLama. Now I'm there and controlling many things sending backend calls to my api connected to my Arduino board, That makes my workflow pretty much easier! Thanks!
This is SO COOL! I can’t wait to try it! If only there was a way to repurpose all my old Echo Dots to serve as satellite units…
They can be used as bluetooth speakers.
How??? @@willdwyer6782
The most difficult part is to buy the things.
Everything else is just a matter of learning the how.
A lot of work went into this video. Well done. This is a real gem.
I have the same pi hat and gave up trying to get it working. Think I threw it against a wall. Going to give it another shot if it still works. Thank you!!
Hey, one thing I did was the use of Assist and GPT in the same command, Assist is waaaay faster to respond to obvious commmand like "turn on the light", but it cannot do AI, so I did a nodered script that use assist and if it fails goes to GPT, it works quite well to be fair
maybe a little hint, how you achieve this?^^
Dude, right on, right on. Use to watch your videos all the time, and now I'm stuck again trying to learn some voice recognition stuff the past few weeks, and you just came in to save the day! Thank you!
22:32 Can confirm that ollama binds its port to 127.0.0.1 on debian 12 as well and you need to configure it to bind to 0.0.0.0 or any specific network interface as well if you want to access it from other machines.
That is about the coolest thing I've seen in home automation, ever.
Don't think I'll do it though. I can copy what you did, and I'm sure it will work. Until some of that software changes. Then, I'll have no idea how to fix it on my own.
I'd love to see links to resources showing how you learned to do all of this stuff.
I'm looking forward to replacing Alexa so I can now spend a few hours a day trying to get my lights to turn on and off
Defender is a great suggestion but I love you approach with virtual browser in the cloud. Your knowledge is awesome your enthusiasm is inspiring
33:35 It might be a good idea adding something like "be concize in the response" to the LLM system prompt so that you do not get crazy long responces read out 🤣
This is amazing you give great step by step description and easy to follow along guide... I am so surprised by the things home assistant is capable of..Please keep the home assistant videos coming
Always a great day when network chuck posts
If you want to build your own hockey puck/snowball, someone already did that, too. I don't recall what all it entails, as I found this article like 2 months ago, but it uses an ESP32 instead of a Raspberry Pi to connect to Home Assistant. Hope this helps someone. Great work, Chuck!
Thank you for muting the drinking sounds, I braced for impact and was hit with delightful silence. It was noticed and appreciated! My Misophonia is evil.
Lack of a ready-to-go voice assistant, and hardware for it, is the reason I've not been able to commit to Home Assistant. Rhasspy may almost have the software, but the cost to build dedicated devices to replace the 8 Google Home receivers around my house is somewhat impractical. However, this video is the best solution I've seen so far!
It occurs to me that you don't need to have a raspberry pi running each individual reciever. Simply set up wireless microphones in place of the various google homes and have them all connected to a central pi.
You're going to spend a bit more than the price of a google home, but a fraction of the cost of 8 of them.
Considering the power consumption is so low you could have the mics battery powered and have them all over your house if you wanted.
It also strikes me that you don't really need them in every room since you can just use your phone 99% of the time. You really only need them in the main living space(s).
33:53 : Bro you made me laugh so much! hahaha i wasn't expecting that 🤣
Best video for this topic until now! I've seen a lot of video tutorials, but this was the more elucidate one. Congrats!
Chuck and gang, I am banging my head against the wall here. I have whisper and piper running on a windows computer, followed your directions but when I go back to add the service, I get connection failed. I have tried everything I know including adding the ports to the advanced firewall settings EVEN though they are turned off. What the heck am I missing here.
Love the video and thank you for all you do. This is going to step my entire house up.
did you do definitely do the localhost step to 0.0.0.0?? I couldn't get ollama to connect without this. no amount of ports etc made any difference for me.
Have you found a way? I am having the same problem.
That is not a Gen 1 Alexa. My daughter purchased me a Gen 1 Alexa for Christmas the year they come out. I still have it running in my bedroom. We moved across country last year. The unit is paired, but it is not possible to re-pair it. Therefore my local weather and time on the device are wrong.
Thank you for this video Chuck. My wife is handicapped and we use Alexa's all over the place and I wanted to replace them with my own setup.
For bonus points, we listen to music with an MPD client/server setup. If there was someway to voice control our local mpd collection that would be awesome!
There's an MPD integration for Home Assistant so it should be possible.
@@eDoc2020 I had seen that if I can get volume up/down play pause, next, previous, what's playing and switch to playlist, life would be good.
Looking forward for someone who gonna make some PCBs to replace some echo devices guts or, maybe, some firmware replacement. The lights, mics and speakers are better, form factor too. All have built-in wifi and some have the screens/displays. Just the cloud/IA stuff that sucks. Taking back the hardware we already paid for would be amazing
This. My dream smart speaker would be some HA-enabled board like esp32 or pico inside a HomePod mini, because those speakers are mind-blowing, they have a pretty clean, full sound with incredibly fat low frequencies considering the small form factor. Maybe even go full Frankenstein and swap the mic for the one that's in Google's Nest mini, this little guy impresses me with how far it can catch my voice from. It would definitely not be cheap tho and quite hard to do, if at all possible. But hey, one can dream, right?
I actually worked with one of the original Mycroft creators earlier in my career. First time I ever saw a 3D printer in real life was at their Alpha launch/Christmas party. It was a fun project to follow along with.
That said, I'm in the process of doing exactly what you did in this video. I'm running Whisper, Piper, and Ollama on my TrueNAS instance, and I'm running HA baremetal on an old i3 Gigabyte Brix.
I have tried doing this and it was so slow and decided not to use it. I had no idea you could off load whisper and piper - GENIUS!! And the Ollama for an LLM - I need to read more. Please more of these if you can!! Thank you!!
One question - instead of a GPU could a TPU work like google Corel Accelerator?
But thank you - more please!!
TPU current RAM is too low. So it won't work.
@@MrThesoulripper13 @trekbear69 current Hailo/Coral TPUs are sadly ONLY good for AI Vision, even just for object detection Frigate BUT NOT for Training, for LLM You need nVidia GPUs, however good news Ollama can now supports newer Intel GPU Iris as well, whisper NOT yet
Awesome video Chuck.. I'm actually in the process of adding features to my HA installation, so I'm really loving these ideas. Especially when it involves replacing Alexa with something that runs locally.
If you eventually decide to containerize Home Assistant and move it to Docker or better yet Kubernetes (would probably make for a good video), just be aware that Add-ons are more limited. You can still install them, but they need to be installed into their own containers and cannot be installed through the HA Add-ons screen, so there's a few more hoops to jump through.
Why would you do that?? Install Proxmox on the machine and make a HA VM, if you need GPU acceleration you can easily just passthrough it... DO NOT use docker for HA...
Can I install all this whisper stuff in a docker? I have Home assistant running on my synology, but I want to replace the alexas with raspberry pi's. Or, should I jsut run it all off the raspberry pi?
@@TheRareGamer what's the problem using HA as a docker? Can I setup this Proxmox on a synology?
@@TuckerDansieFilms Because its more of a PITB for doing add-ons in docker
Though I do run it in docker myself - but I also don't really use add-ons...I think I use one but then I might also have just given up too
I think you meed to run hassOS to use pluggings.
Just wanted to let you know i love your energy when you make your video's, something a lot of youtubers lack! Keep it up!!
I am having some trouble if anyone can help. I'm at about 26:03 (ish) and I can run the docker command fine and the whisper service is running> I get the INFO:__main__:Ready but when I go to put in my IP and port 10300 it cannot connect. I see my WSL is running a different IP than my PC but I don't know enough about all this to understand why Home Assistant can't find my whisper instance running on port 10300 when it's active and running. Does anyone have any suggestions?
Same problem here, we need help pls
@@torouxFR Hi! anyone found a solution? i tried putting land mode on "mirrored" but no luck
GPU acceleration with docker whisper is pretty cool. Though Llama3.1 70B eats up so much of my VRAM that it does not fit. It enabled you to run the medium-int8 model which is a lot more accurate and fast with GPU. I'm currently trying to get AMD GPU acceleration to work with my Proxmox with AMD GPU on a Ryzen 7 4750. This is the coolest project ever. My voice assistant mocks and insults me relentlessly. I'm so happy!
5:49 NOOOOOOOOO DONT REMOVE THE COFFEE SOUND
EXACTLY MY REACTION!!!!
Totally agree!!! Stand up to the anti-slurp tyrants, Chuck! Don’t let anyone bully you into self-censorship!
I didn't actually notice this the first time, but now that I know, I can't unknow. Why would they ever remove the sound?!
Network chuck video isn't complete without the iconic sip ...
I love the coffee sound, I must be weird 😂
Very cool Chuck, I am just now getting started with your academy (the free version) I wish I found your videos sooner nonetheless it is a great time to be alive. I am an electronics technician retired USN my interests now are in Ai and learning IT to become a network engineer or at least an aspiring one. I will be running my own corporation as well as it is just now starting up. A lot of what I have learned in my time (just being alive) is I dont need to know everything to be a leader in any field just a good headhunter. To do this I must know more than the basics and for this type of business my mind is like a dry sponge. Your videos are inspiring and easy to follow. I have come up with my own ideas about practical use of AI and now I see many people have already been working on this. I am encouraged as most people still dont know this is the perfect time to be in IT Networking and AI development. Thanks to you my journey is becoming much smoother. Thanks and I will be subscribing and following you on youtube.
I'm in the Apple ecosystem and have been using HA for years. Their homebridge is incredible and actually allows me to just use siri to control everything I expose to it. Devices, scenes etc.
Your art of doing These Videos ist exactly what i'm searching for. Very nice, Go Go
Thank you very much
You can run Ollama as a docker container now!
What does this mean? How?
yeah but u don't need to
I wish that when I attempted college for computer network and security my teach would have been like you.
This is SO amazing! Congrats Chuck, you are at the same time smart and funny! Love it! Hoping for the Video #2 to hear Terry Crew's voice!!! 😀
Can you help me please.
My Gmail account was hacked.
So I want to back it by hack again.
I don't know hack 😥😥🙏
Thanks!
I love what Nabu Casa are doing with Home Assistant. I actually really loved the telephone gimmick they have where you can use an IP Phone to talk to the voice assistant. I know most people wouldn't use it, but something about picking up a phone to call home assistant was kind of fun.
thats interesting. if u can help with some links or something it will be helpful.
also i have a question if u know
Im new on this world :)
I have a raspberry pi 5 / 8gb Ram and installed the HA on an sd card 512gb U2 class 10. When i installed this addons all the system crashed and a message appeared that is out of network. I do this again and again and the same msg appears do u have any idea what its going on ? i know u cannot help me with this info but am asking for an opinion if anybody knows.
@@martinospapantoniou4491 UA-cam is very good at deleting comments that have links etc. I hate doing this, but it's the best I have. Do a quick search for voice and home assistant. :) Wish I could give more info.....
i loved watching you deal with it running over you with long messages. it happens for me often
so. very. often. ChatGPT, please stop talking and just say yes or no when I ask you something.
Dude chill with the caffeine or you're gonna end up being a guest on a James Hoffmann UA-cam video with a moka pot in your hand.🤣
Chuck you can make aliases of your devices in home assistant and call the what ever and expose them to your pipeline. It's kind of a pain in the butt and a long process, but at least you could call your lamp whatever you wanted to when you talked to your voice Assistant
What has the YT algorithm done to you? The over excited gestures and speech that you see in ipad kids. Fight it don't let the numbers change you, its not to late.
Bro he has always been like this. What are you smoking
So happy you started doing HA and AI tutorials, thanks 😊
First
Sadly ollama (AI in general) needs lots of power to work Alexa like. For someone like you, who has a power server to host that, thats nice, even a bot on telegram to have your own AI.
But for the common guy with only a raspberry thats not doable (yet)
Glad to see it working fine (as long as you can provide resources to it)
One of the commands we use with Alexa is to start timers. Sometimes we have multiple timers running at the same time. Can you show an example of starting a 2 minute timer, along with a spoken phrase of "your timer is done" (or "your egg timer is done" if you've given the timer a name of egg). We tell Alexa to start a timer by saying "alexa, start a 5 minute oven timer". It is fine to have Terry announce completion once and turn off, as opposed to alexa that continues to alert until you tell it to cancel. Also need a way to cancel a timer in progress.
Like most home assistant projects -- it looks like a lot of work, but what an awesome result.
you have alot of pc skills...much impressed...I'm sure I would not make this work....i'm having trouble getting home assistant to run on a mini pc.
"I don't care how much coffee it takes; I'm going to do it!" LOL This is the second video of yours that I have watched, and I now understand why you have over 4 million subscribers. Pretty amazing stuff. You really know your coffee. Keep up the great work; you won another subscriber. I can't wait to set up the voice assistant on my HA system.
This is Your BEST video! Cant wait for the second part....Good luck
Great to see that you are such a big fan of Home Assistant NC!
I like how he says that he does not want some other party to have his data, but 80 percent of his other videos are building something that can run flawlessly on a local machine, but he makes it online on some random VM somewhere in the world :)
I'm speechless. I've discovered your channel because of this video and I can't even follow your speed just watching you writing code XD. I'll try to follow your steps just to achieve the dream of having my own "BT-7274" as an assistant. Thank you so much for shedding light on this topic and I'll be waiting for your "Voice change" tutorial. I've already extracted the BT voice lines from Titanfall 2 but I wasn't sure of some steps explained in the documentation. Hopefully, you will know how to do it. And again, thank you so much for this content. New sub!!!
OMG. Thank you for doing the work. This is great! I'm going to do this in January. You've made it so much easier.
Holy Rabbit hole batman!! Mastered Stable Diffusiuon in the last few months and now fell into LLM home automation
I have been meaning to do exactly this from a long time. Boy oh boy, you saved so much research time.
Two needs our alexa ist most used for:
- Setting timers
- Playing radio stations or other music from Amazon Music / Spotify...
Man, I've just bought HA Yellow , and got to look into this vid - awesome. Gonna bookmark it and implement tomorrow. Planning to run Ollama thing on my Z420 server, in a separate VM.
You Sir have my respect.
Something you might try to add to these videos using docker is using compose or even portainer. Makes managing docker in general easier, especially if you are doing multiple containers as it helps visualize things a lot better.
0:26 - The scent that never dies
I was just attempting this on my own, but now I have some more experienced guidance lol, This is so cool!
DAMNIT U TRIGGERED MY GOOGLE!
I said "cancel" and it closed this video on my chromecast 4k😂
Hey Chuck! 👋
Shouldn't it also be possible to accelerate everything with the Raspberry Pi 5 and the AI HAT+ if you don't have an AI workstation? That would be interesting to know.
that "I'M GOING TO DO IT" at the end was so funny 🤣 can't wait to hear Terry's new voice!
34:10 Would have been super useful to have a side-by-side or timed comparasing of the 3 setups! all on raspberry vs offload to laptop vs offload to terry. That way you get an idea on what is worse case, average and best case scenario! A followup would be nice.
My brain just melted when I just noticed I could possibly re use the speakers I don’t use on my 3 seater excalibur / melendez couch. One for the pi / mic, one for the speaker
Totally awesome. My home assistant is already running on docker. Let's hope that my good old QNAP is able to run ollama. But since HA addons are also docker containers inside HAOS, you just need to run your HA on a better HW like a mini pc
Love your work Chuck.
You mentioned that you don’t use frigate. I would love to see it though with home assistant and local AI, when object, person detected etc it send an alert with AI generated description of what it sees
Chuck you are amazing!!! I love watching and learning from your videos. Please keep them flowing.
This video is awesome. I spent way too much time trying to offload all of the AI stuff to my other server only to get lost in docker stuff. Thanks for putting this out there. Also, getting a custom voice setup is nearly as fun as removing a foley catheter while its inflated. I was ultimately successful, but the voice sucked and the juice really wasnt worth the squeeze. My hope of getting into an argument with Dagoth Ur, the god, have been postponed.
seriosly man i was searching for this since 1 month , thanks for making
I thought you were going to have some servios closing the doors and raising and lowering the toilet lids In your home assistant video ..but nice vid anyhow lol Love the video!
This is absolutely wow! Would you consider doing a special "upgrade" of Terry incorporating an n8n AI pipeline with more datasources, local folders/databases and things using Home Assistant? Is this even possible? Would probably be like putting "something" between HA's conversational agent and Ollama, right?
This was fun to watch! Thank you, Chuck
1:21 That's the first widely available version of the dot, but it's the 2nd model overall. The very first was only available to purchase using a voice command with the original pringles can sized echo. It was all metal and only had the wake and mute buttons because the volume was controlled by twisting the top. They were pretty awesome, but they sounded like crap and they were inordinately expensive compared to what they charge for a dot nowadays. I paid $80 IIRC.
This is one of the coolest tutorials I’ve ever seen. Can’t wait to see if you get terry‘s voice set up.
Noice! Great timing btw as this topic is my very exact winter project for this year!
Super excited if he gets the Terry voice going. I've been wanting to train mine to sound like a transylvanian lab assistant. I named him Igor. Id like to get as close to Marty Feldman's character from Young Frankenstein as possible.
I think Terry should use a Borg voice :)
Good luck with changing the AI voice. I was watching the STNG episode, "Skin of Evil," and thought the voice of Armus would be funny for an voice assistant.