my local, AI Voice Assistant (I replaced Alexa!!)

417.51k views8683 WordsCopy TextShare
NetworkChuck
Scam Copilot is available for new users of Bitdefender Premium Security and superior plans. Find out...
Video Transcript:
I'm done I'm throwing Alexa in the trash I want to create my very own fully local AI powered voice assistant and I'm going to go a bit crazy first I wanted to use AI because Alexa's kind of dumb but I don't want to use AI in the cloud I'm tired of people being all up in my business get out of here Jeff so I want to do everything local in fact I want to connect it up to my AI server Terry that I built in this video here in fact I wanted to answer to Terry
and have the voice of Terry this voice I have a s can we do it I I don't know okay Nabu what year did The Matrix come out I hate you I hate you so much okay Nabu what year did The [Music] Matrix I quit I quit if you do want to follow along with anything I'm doing here here's my parts list starting with level one basic and easy and then here's all the stuff I have for advanced everything will be below now it turns out this is actually not that hard to do at least
at level one level three and four are kind of insane and that's that's where we're going that's kind of where Terry needs to live and by the way I'm showing you this journey as kind of a tutorial so if you want to follow along and build this yourself let's do it again the first one's not that hard but the last one I think it'll be level three or four will involve training a voice with AI which took me way longer than I expected and I had to rent a GPU server on Amazon we'll get to
that now as I said before I've been using Alexa for a long time this is like one of the first alexas I think it is the first one and by the way I'm so sorry if I'm setting off all your stuff even more of a reason to go local right but seriously I'm done with cloud-based stuff I want to go fully local for everything in my last video I started that process by installing home assistant on a device like this a Raspberry Pi home assistant is a fully local home automation software and it's fully free
and fully awesome so if you do want to follow this tutorial and set up your own local voice AI assistant go watch that video and see how to set up home assistant because you will need that why I'll go into that here in a moment now replacing your home automation that's actually not too bad it's pretty easy but replacing voice harder to do mainly because there's not a ton of options out there at least not options I can meet my laundry list of what I want to do now my first task was to see who
has gone before me has someone already done this the answer is kind of and then they gave up but now we kind of have something a little bit but it's hard to figure out let me explain I found the perfect solution it's called mcraft they even had a device like Alexa take my money but okay it's dead discontinued it was killed off by a nefarious patent Troll And even their Kickstarter backers were left empty-handed they were supposed to receive one of their devices and they look pretty cool I mean look at this thing that would
have been awesome and beyond that there really wasn't much I was coming up empty but then for some reason I overlooked home assistant wait what you mean the thing that's controlling your home yes I like you thought okay that's just my home automation now I need a voice assistant no they just had their year of voice in 2023 and their goal was to implement voice control for the home okay sounds pretty interesting okay what they've done is way more than interesting check this out they enabled support for an open source tool called raspy which I
had never heard of until this moment moment a fully offline check set of voice assistant services for things like home assistant now along with having the worst logo I've ever seen what even is that they do things like open wake word it's just built in and that's being able to wake up your assistant based on a voice command kind of important so think Alexa Okay Google Terry it'll do St or speech to text again this is all running locally on your hardware and you actually have options for what you can use I went with using
open ai's whisper and not the cloud stuff the locally hosted model they provide and then they have the reverse because you want your voice assistant to talk back to you that's TTS again a few options but I fell in love with this tool called Piper now raspy is not new but what is fairly new with home assistant is the intent the intent recognition when you tell your voice assistant to do something like hey turn off that light or what's the temperature of the bathroom right now and needs to take what you say and figure out
okay how do we make this a command for the system to actually do so home assistant came in clutch with their thing called assist which which allows raspy to talk to assist and tell it to do things like turn on and off lights turn off the internet and that by itself is cool but home assistant did one very cool thing that I didn't see an easy way to do anywhere else you see I wanted raspy which essentially is just a voice tool it doesn't have any like smarts per se it's not like AI it's not
like talking to Chad BT so I wanted to have raspy talk to an llm similar to Chad BT so than ol Lama with llama 3 I couldn't find an easy way to do that but home assistant makes that stupid easy they actually have haveit to where an llm like AMA can be the brain and when I realized this I went and tried it and it's kind of easy to set up let's set it up right now you don't need anything special all you need is the device you're running home assistant on which I have it
running on this Raspberry Pi right now again if you don't have that set up go watch my video now I'm starting from scratch so I can show you everything from the beginning so here I have a fresh installation of Home assistant I'll add one device so we can play with stuff and by one device I mean all my Phillips Hue Lights so lots of devices and real quick what we're going to do is set up what's called a voice pipeline it sounds way scarier than it actually is just a few steps so here in home
assistant we'll go into settings we'll click on add-ons which will add additional functionality to our home assistant system home assistant system kind of a tongue twister then we'll go to add-on store and we're going to add two things We'll add whisper for our stt or speech to text so we'll just search whisper there it is Click install and this will actually be running whisper transcription locally on your Raspberry Pi and if that doesn't make you excited what does what do you do for fun this is what I do for fun and drink coffee once installed
we'll click on start so whisper takes our speech and makes it text now we need something that can take text and make it speech a TTS so we'll go back and search for one more add-on it's called Piper Piper is kind of amazing also kind of frustrating you'll see why here in a second click on Piper click on install little coffee break then we'll start that sucker and by the way this is all level one stuff super easy and then we're going to add two more things we go back we'll search for assist microphone this
will allow us to use the microphone of our our local device and actually I have this Blue Yeti microphone I'm going to use can really be any microphone you want I know it's a mi error but still I'm going to go and plug this in by the way you don't need a microphone to do what I'm doing red light we're good to go so I'll install that and before I start it if I go to configuration at the top here and scroll down just a little bit until I see audio I can choose my options
for input I'll say My Blue Snowball mono and actually you know I have a speaker too I'm replacing Alexa with this Behemoth all right let's see if it finds that speaker now how about I save it go back see oh it even saved my settings there it is save and now let's start it and then one more add-on we're going to add open wake word click on that install again this will enable us to be able to wake up our assistant with a command or a name and we'll click on start add-ons are done from
here we'll go to settings and then devices and services and notice they autois discovered two weird things these buffalos actually a bunch of weird things or are they bison I don't know what are these bison doing on our screen this cattle this is the Wyoming protocol we'll go into this a bit more here in a moment but just know now it's the Wyoming protocol that discovered our add-ons so whisper Piper open wake word and our assist microphone all we have to do right now is click on configure on each one of these submit finish configure
submit finish configure submit finish last one configure submit finish now would you believe that we're almost done with having a local Voice Assistant check this out all we have to do now is go back to settings and then go into a special place called voice assistants now they already have one there by default just called home assistant let's edit that one click on that and here we're kind of configuring our voice Pipeline with the components we just added we'll scroll down a bit until we see speech to text here we're going to click that drop
down and we should see faster whisper that's what we just installed for St and then for text to speech we'll change home assistant Cloud to Piper we also just installed that and then for the Wake word we'll choose our wake word engine we just installed that and we'll choose our wake word Alexa no we've been saying that way too much let's do microft rest in peace and update but what should happen is I should be able to talk to this I think let's see okay micraft how are you today nothing why isn't it working let's
go troubleshoot make sure our configuration was good yeah let's try restarting some stuff okay Nabu how are you that scared me how are you sorry I couldn't understand that that was terrifying fine all right I've got a light behind me let's see if you can find that light now what's it called Chuck lamp light okay Nabu turn off Chuck lamp light sorry I'm not aware of any device called Chuck Lamplight now first of all that's just cool second why isn't it working I know why okay Nabu turn off Chuck lamp light in Studio One turned
off the light now notice I had to be very specific and it was a bit slow we're going to change that in the future levels comeing out but how cool is that all running on a Raspberry Pi that's powerful now you can also chat with it right here in your browser by going to the top right and clicking on that assist icon and saying turn back on Chuck lamp light in Studio One turn on Chuck lamp light in Studio One got it now I can't use voice right here because it's not using an https connection
browsers block that if you do want to try the voice and you don't have anything to try it with install the mobile app you'll be able to talk back and forth it's pretty neat hey real quick can I show you something awesome watch this okay Nabu who is the sponsor of this video Bit Defender is the sponsor it is the best thing ever okay I'll tell you how I did that here in a second but watch this okay NAU what does Bit Defender want me to talk about listen Chuck they have a set of new
features in their platform specifically scam co-pilot go check your notes and stop being lazy how cool is that but seriously the sponsor of today's video is Bit Defender and yes scam co-pilot is a new set of features part of their new platform and is included in their all-in-one security plans now by far the coolest thing is what they can do on your phone watch this I just installed this first of all I can chat with my scam co-pilot right here then it can also do this and I had no idea this was a feature if
I go to finish setup for full protection it can do SMS protection what so I go to my set and I go to messages here in unknown and spam message filtering I can specify Bit Defender to filter my messages for me and use the power of spam co-pilot yes heck yeah I enable it I can do calendar invite protection that's cool I mean how many of you have gotten these random weird calendar invites you're like what is that probably spam okay Nabu where can they learn more about scam co-pilot Chuck drink some more coffee you
already know this they can learn more about Bit Defender scam co-pilot by clicking the link in the description don't worry it's safe we checked it you heard her now how did I do that and by the way thanks to Bit Defender for sponsoring this video and being a continued sponsor of my Channel please show them some love and check out their stuff below but watch this this is so cool and we're just scratching the surface if I go to settings and I go to automations this is how I did this I created a few automations
that respond to a certain voice sentence I say voice sentence so for example the first one was new automation the trigger is when I say something like who is the sponsor of this video and then I can have it respond with a certain response and that's how I did all these things now you can do a lot more than that but we got to move on okay level one done but here's the thing if we're going to replace Alexa we got to have more than just one right and it can't just be running on the
home assistant device itself we need others that we can use and connect back to this this is where the Wyoming satellite and the Wyoming protocol come in now the Wyoming protocol our beloved bison Buffalo thing I'm not sure what that is it's is a small peer-peer protocol for voice assistant now what does that mean for us it means that we can take a Raspberry Pi for example which I'm about to do that right now with this guy right here ah he's kind of Wiggly and we can make the Raspberry Pi a Wyoming satellite essentially a
remote Voice Assistant that can communicate with and use the voice pipeline from home assistant now this would be kind of level two not too scary but it does involve some command line let's get going you might want to drink some coffee for this now if you want to do the whole USB plugin thing you can you can get a snowball ball and a a giant speaker and you can have that with a Raspberry Pi plugged in all around your house you can also use something like a Raspberry Pi zero there's a video from this guy
right here who shows you how to build one I didn't happen to have one that I wanted to use but I had a bunch of spare raspberry pies now a very quick and easy way to give your Raspberry Pi the ability to hear or a microphone and the ability to speak or speakers is with this thing right here where' it go I've got one it was in our inventory list below this is the ree two mic piie hat this little guy has two microphones on it and connectors for an external speaker now I'm going to
unplug my pie real quick and pop on this hat onto its IO connectors these are Pretty stinking cheap and they do have LEDs on them so you can give a little status light for when you're talking to your assistant so I'm going to plug it back in here now this is a fresh Raspberry Pi meaning it has an SD card with a new image applied to it I used for my Raspberry Pi I'll show you real quick it's a Raspberry Pi 4 with I'll go to Raspberry Pi other and it's the Raspberry Pi OS light
64-bit that's all you need so make sure you image the SD card pop pop it in and if you need help with setting up Raspberry Pi I do have other videos showing you how to do that so go ahead and launch your terminal this will be the same for Mac Windows Linux and get logged into your Pi here I'll do a pseudo hbt update and I'll have all these commands below first I'll install git and Python 3 virtual environment just like this and then using git I will clone the Wyoming satellite repo from raspy it's
actually run by raspy then I'll jump into that directory the Wyoming satellite CD Wyoming satellite and this next command is only for you if you're using this little hat we have the re speaker 2 mic or for mic hat we're going to recompile and install the drivers with this command the drivers are telling a Raspberry Pi how to talk to this guy oh and I forgot to mention I've got a little bitty speaker that I'm going to plug into our re speak pie hat it sounds like not great but it's still pretty cool I love
how tiny it is and then we'll do a pseudo reboot we'll get logged back in when it comes back up all right now that we're logged back in we're going to CD into Wyoming satellite and we're going to create a python virtual environment with python 3- mvnv VNV these are super powerful they help you isolate things and not step on anything else's toes and then we'll use these few commands to upgrade pip which is a python package uh manager installer then we'll use wheel to do some setup tools easy for me to say and then
we'll run this very long command to install all the requirements again all commands are below all right now we're going to run one command to make sure everything's working it should give us a help file if it's good so if you don't see this but if you do see this you're good now we're going to look at our audio devices now I have the re speak P hat I've installed drivers to make make sure it can talk to the Raspberry Pi or vice versa if you're not doing that that's fine run this command a record-
l capital L to see your audio devices find the one you're going to use I'm going to be using this one right here but if you're plugging in some headphones or a Yeti mic just keep that in mind right now want to make sure everything works I'm going to run this test command again fill in your specific mic that's going to record something right now from this card for 5 seconds and save it to a test. wve file let's do that right now I love when stuff like this becomes practical hey Network Chuck here I'm
going to take a sip of coffee right now now let's play that back with this command a play ready set oh wait I forgot to put my speaker in oh no let's add it real quick there we go and playt Chuck here I'm going to take a sip of coffee right now that's so fun okay speaker working the Wyoming satellite is set up let's test it real quick we're going to run this script copy and paste this is very specific to this Reb hat if you have something else make sure you exchange these commands here
with your device now what this is doing is going to run the satellite on Port 10700 and that's how our home assistant running over here is going to connect to it let's run it right now it is running excellent now let's go set this up in home assistant we'll go to settings devices we'll go into our configured section because we already have Wyoming configured and find Wyoming there it is Wyoming protocol click on that and then we'll click on add a service here we'll put the IP address in of our uh Raspberry Pi gosh what
was mine again I forgot and it doesn't have to be a Raspberry Pi you could have been using another device you can do that and then Port 10 700 now fingers crossed we click submit it found it we'll leave it in default area click on finish it's there there now right now that's all we had to do our voice pipeline's already in place we added this via the Wyoming protocol it should just work okay Nabu how are you found sorry I couldn't understand that yeah I think it does get a bit weird I'm going to
it's like I can't understand that I'm speak in English I'm going to click on the Dots here and say reload the satellite let's go watch the command line see what's going on disconnected from server and we're back in so now it's waiting okay Nabu how are you sorry I couldn't understand that okay it's being kind of dumb we're going to fix that dumb problem here in a moment with level three okay Nabu turn Chuck lamp light off turned off the light ah we got it okay you had to be very specific okay Nabu turn it
back on sorry I'm not aware of any device called it back now you see that that's stupid and one thing we're going to fix you'll be able to have a conversation with it it'll remember the context W it's so awesome so we verified it works but now we got to turn this Wyoming satellite on the Raspberry Pi into a service so it runs all the time so when you unplug it and you plug it back in it's always going to work so stop this by doing contrl C and we're going to run this command pseudo
system CTL edit Force full Wyoming satellite service we're actually creating a service on this and we're going to paste in this config keep in mind you're going to want to make sure you change the user so right now it's Pi which is incorrect for me it's got to be Network Chuck so change it to your user is going to be referencing those directories and also keep in mind it's going to be using the command that's very specific to my pie hat change it to your appropriate audio card contrl X Y enter to save we're going
to use system CTL once more to enable that service with the enable command man now let's restart it oh no she already be running let's do a system CTL status to make sure it's going looks good from here okay Nabu turn Chuck lamp light on turned on the light I mean that alone is pretty cool but not cool enough we're going deeper now one thing I hate is that the there are LED lights on this but it's not using them I want to be able to know like when it detects I'm talking to it let's
enable that real quick that's going to be another service if you're not interested in that move on so to get our LED with the re speak device to work run this command to install some python libraries and then we'll CD into the Wyoming satellite directory and the examples directory oh we just got to jump into examples we're already here then we'll create another virtual environment this time just for our lights and then we'll do some basic installation stuff so upgrading pip setting up wheel the whale installing Wyoming 152 and then finally installing the pixel ring
python package for our LEDs now let's test it with this command testing out two mic service and if you saw the results of a help command were good now we're going to create a service to run that led program all the time so when this command system C CTL edit it'll be the two mic LEDs service we're going to paste in this config changing our username here to contrl X Y enter to save and then we will have to edit the Wyoming service config file and add some info about those LEDs adding this config requires
two mic LED service so it's going to be dependent on that and at the very end of your exec start command you're going to add this bit event URI TCP 10500 that's the port that your two mic LED is going to be running on contr X Y enter to save let's enable our two mic LED service with this command and start it changing enable to start let's check the status so far so good looking good let's restart our Wyoming service and yeah we got some lights okay let's test it out it should change when I
talk to it okay Nabu how you doing sorry I couldn't understand that all right so now that's all we had to do our remote satellite is configured calm yourself it's connected to home assistant and we can just take this and unplug it in anywhere we can create multiples of these and they're all going to be using the Wyoming protocol to talk now time to get more insane I think we're on level three now if we go back into home assistant and go to a voice assistants once more and we click on home assistant you know
what I don't like first of all this conversation agent not smart enough for me it's too dumb home assistant doesn't know what's going on I want that to be o Lama I want it to be a llama 3 actually I want it to be llama 3.2 a very very smart local llm that you can run on pretty much anything you want you know what else I want the speech to text whisper I want that run on other Hardware as well because this Raspberry Pi right here it's Raspberry Pi 4 not very fast I can actually
run this whisper on another server now what I'm going to do right now is demo running it on this laptop behind me if you can see that can you see it you can now so I'm demoing it on this laptop but I'm ultimately going to use Terry I'll do the same thing for Piper too I'm going to offload everything that's AI intensive to another server let's get to work okay here on my laptop I'm going to install AMA so I'm going to go to the Internet find AMA ama.com go to download and what I'm showing
you here will work for everything click on window Windows download did it download oh yeah it's downloading up there we'll run the setup now obviously if you got Mac or Linux click on those the installation is actually easier than this now what Alama allows us to do is run local llms or local versions of chat GPT except they don't touch the cloud no internet they're amazing perfect for new Terry coming up soon okay it says AMA is running click here to get started Aha and look it gave us the exact command I want to run
AMA run llama 32 my favorite model right now so we'll run that Command right now it's going to download llama 32 and it's going to run it real quick I say quick like a minute okay it's running we can do a little test like hey and I I don't want any assistance so I'm going to do buy is it buy yes buy okay it's running now we can actually test that by going to our web browser typing in Local Host and then Port 11434 awesome ama's running that's how home assistant is going to connect to
us or the AMA server so now we're going to go back to home assistant we'll go to settings and then we'll go to devices and services and then we're going to click on at the bottom right add an integration cuz the new integration we're going to add is olama it's official it's supported it's awesome click on that and then search for oh Lama there it is click on that and it'll put in the IP address of our laptop which mine was 10. 70.7 do or no 10. 72104 and they'll put the port in of 114
ah I can't type today 114 34 fingers crossed H why okay I think my Windows Firewall is on let's go turn it off that will definitely block stuff well it's turned off okay I know what it is Alama by default runs on 1271 we need that to be 00 00 on all interfaces not just our loot back port to allow connections from the outside so annoying you probably won't have this issue on Mac or Linux I'm going to quit a llama Biol Lama we'll search for environment variables and we'll say edit these system environment variables
and then click on environment variables down here under system variables we'll create a new one called _ host and the value will be 0.0.0.0 and that's it now let's launch AMA again llama it should start running let's check it in our browser it's up and now let's add it to home assistant man I hate that little detour we had to make one 434 Ready Set stinking yes all right let's choose our model I'm going to use 3.2 the one we downloaded and click submit ah it's there so now how do we use it with our
voice pipeline let's go to settings this is going to be fun go to Voice assistants jump into home assistant and under conversation agent we're going to change it from home assistant to llama 3.2 now one thing we do have to do if you wanted to have access to our stuff in our home we have to click on that gear and then go from No Control to assist to give it access you can also kind of tell it who it is actually I like to add cuz it might not answer you it might not answer questions
you ask you say you can also answer any other question you have options here but we're going to ignore that for now click on submit finish and update so now we're using AMA this is this is so fun check this out we're going to go to uh we'll go to our laptop and launch task manager so we can see our GPU go crazy when we ask it questions okay here we go okay Nabu who was the first president of the United States there it goes it's using it the first president of the United States was
George Washington he served from 1,789 to 1,797 and is often referred to as the father of our country yeah did you see that Spike what's really cool is the context okay Nabu turn off the light Chuck lamp I have successfully turned off the light in Studio One okay cool now let's see if it keeps the context this is what we could not do before okay Nabu turn it back on yes I was able to turn on the light Chuck lamp okay perfect okay so now we're using a local llm hosted on my laptop back here
but honestly things are a bit slow I don't like that so the next phase the next level in our plan our mission we're going deeper we're going to try and host more and more of that voice pipeline off of the little Raspberry Pi no it's this guy this Raspberry Pi on to more beefier Hardware like my laptop here let's go to our voice pipeline real quick back here in home assistant we'll go to settings we'll go to U voice assistants and click on home assistant the first thing is our speech to text faster whisper I
want to offload that and host it on that laptop let's do that right now here on my Windows laptop I'm going to launch terminal and this will be a Linux based thing it's going be actually a Docker container so if you're running Mac or Linux you're good to go if you're running Windows you'll just need to do some wsl2 magic now if you're like Chuck I don't know what that is I've got a video right here check it out I'm going to launch a fresh WSL instance onall Ubuntu 2204 create some new login stuff and
we'll do a pseudo AP update to update our repositories if I put my password in right I always mess that up and then let's make sure we have Docker installed we'll do a little pseudo AP install docker.io and then right after that type in Docker compose because we'll be using that compose is amazing you're like Chuck I don't know what that is I got a video on that too right here we'll do A- y at the end and install all that and actually I lie we're not going to use compose but hey we have it
installed now real quick we're going to make one directory mkd we'll call it whisper data just like that now with one command we're going to launch a container called Wyoming whisper this is also run by the raspy open source company and we're going to change one thing that's the path because we're doing a volume mapping here in Docker we need to change the path to that folder we just made so we going to take all that out and just put in Tilda SL whisper data that should do it oh and we got to do pseudo
Docker run Ready Set fingers crossed let's go it's going to download that image and notice this is Wyoming whisper not just whisper it's whisper with the Wyoming protocol built in and it's running so you see right here info main ready now we need to add this to home assistant and because we're running the Wyoming protocol it's going to be so easy watch this it's ridiculous here in home assistant we'll go to settings we'll go to devices and services and we'll scroll down until we see our Wyoming protocol right here and before we do anything I
do want to rename our old whisper because it might get a little confusing yours might be called faster whisper if you just did this in a similar way that I did but we can click on the Dots here click on rename call it whisper old and we'll actually click on the one entity and the faster whisper here and we'll change that name too go to the gear and change the name up here to faster whisper old if there's nothing there just put faster whisper old up dat now we can click on add service here we're
going to put the IP address in similar to what we did with AMA the IP address of our machine gosh I forgot it stinking again 1072 104 need to get that tattooed and the port was 10300 right yes 103 or 10300 I'm losing my mind fingers crossed click submit success we got it now all we have to do is go to settings go to Voice assistants jump into our home assistant and then scroll down to speech to text and change that from faster whisper old to faster whisper that's our new thing and click on update
now your Wyoming satellite or whatever you have may do something fun here real quick it'll refresh go red and blink and it's back now let's see what happens okay Nabu what year did the movie The Matrix come out the movie The Matrix was released in 1999 that was pretty fast I got to do a comparison let's change it back okay Nabu what year did the movie The M come out the movie The Matrix was released in 1999 now just so you know I'm not sure I mentioned this earlier in the video you don't need any
special device to be able to use your voice assistant that we're creating now you can actually use the home assistant app which is free like everything free and open source and you can just click the dots to the top right go to assist and have a conversation what year did the movie Gone With the Wind come out yeah there there it is using my stuff I couldn't find any information on a movie called gun with a win could you please provide more context or check if the movie title is correct forget about that tell me
how many movies Tom Cruz has been in now the playback is slow Tom Cruz has appeared in over 50 films throughout his career including Top Gun 50 films okay so that's our next task actually we've offloaded the speech to text now let's offload the text to speech the TTS actually before we do that let's close the stalker container and let's run it as a Damon so over here we'll do run-it that way it's going to run all the time it's not going to just be sitting here do a pseudo Docker PS there it is oh
and by the way this container right now is not using our GPU now this laptop does have a 3080 a notebook quality 3080 but it's not using the GPU it is possible to get the GPU to work with this Docker container but I'll tell you one thing it's not stinking easy I tried for hours and hours and hours I finally got it to work on Terry could not get it to work on him honestly I'm not sure it's that worth it but if you're crazy like me you'll you'll you'll push through but no GPU is
really easy I'll show you how Terry does in a second now let's get rid of Piper on the home assistant device let's offload that text to speech so we'll get back to our laptop and again Windows Mac Linux this will all work on everything this will be another yet another Docker container we'll create one more directory inkd we'll call it Piper data and then we'll paste in this command another Ducker run command and we'll change one thing just as we did before the path to our Piper data excellent now we're going to run it interactively
to make sure it's working Ready set go it's going to pull down the oh wait we forgot pseudo PSE sudo it's going to pull down the docker container and it's ready now next step let's add that to our home assistant our voice pipeline what port was that again Port 10200 so we'll go back to settings devices we've been here before go to Wyoming protocol and just like we did before let's rename Piper so we don't get confused because it will use the exact same name say Piper old and we'll jump into the entity and rename
that as well and update all right let's get back add a service put in the IP address of our server I think I remember it this time 104 got it and then our Port 10200 Ready Set fingers crossed go Piper was found awesome click on finish now let's add that to our voice pipeline go to settings voice assistants home assistant scroll down just a little bit until we see text to speech we'll go from Piper Old to Piper our new one and click on update I'm so curious to see if this is any faster we'll
let that refresh done and let's see if it actually works okay Nabu what year did Bill Gates create his computer Bill Gates co-created the first successful version of the operating system msdos in 1981 and his first commercially successful software product basic for Microsoft was was released in 1975 okay Nabu how long after that did Steve Jobs make his I couldn't find any information on a computer created by Steve Jobs however I can tell you that Steve Jobs co-founded Apple Inc in 1976 with Steve wasak and Ronald Wayne and he played a key role in the
development of the Macintosh computer in the 1980s okay now that's just awesome we've offloaded pretty much our entire actually no we offloaded everything our entire voice pipeline onto our computer here it's no longer running on home assistant and similar to the Wyoming whisper the Wyoming Piper container can be run with the GPU I'm just not going to show you how to do that because there are too many individual things I would never be able to show you it and it it probably would never work for you I'm just telling you if I showed it to
you you have to dig into it yourself CPU actually works pretty good now let me show you Terry I'm going to add Terry to do this I'll simply add that in my Wyoming protocol putting in Terry's IP address faster whisper and Piper and tell you what let's add uh let's add his chat GPT as well through AMA so it's all running on Terry yes you can add more than one llm from various servers or the same server if you want I'll just pick one of my llms here actually I need to rename this so I
don't get confused we'll do llama 32 done let's go adjust our voice pipeline I think it's all ter right now yep click on update okay Nabu how many Batman movies are there I don't don't have a specific answer to your question about the number of Batman movies could you please specify which movies or era you are referring to is it the DC extended Universe DCU or the previous films based on the comics oh wait I didn't give it to Terry I give the L on the Terry update I totally did not do that okay let's
try it again okay Nabu how many total Batman movies are there across all the universes there are numerous Batman films including one DC extended Universe DCU asterisk Batman v Superman dawn of Justice 2016 asterisk justice3 animated film asteris Batman and mr9 animated film asteris Batman of the batman3 Harley Quin 2020 wild I asked the wrong question is listening all the Batman movies univers and mediums here is the revised list one DC extended Universe DC unplug that that was funny oh boy let's plug that back in okay here's the state of things we've gone pretty far
we now have a voice assistant that's using some serious Hardware Terry my AI server for all of its processing from speech to text text to speech and it's using the local llm llama 32 and it can control my stinking house okay Nabu turn off Chuck lamp the Chuck lamp has been turned off it's Pretty stinking fast okay Nabu turn it back on the Chuck lamp in Studio 1 has been turned on do a short flash on Chuck lamp the Chuck lamp in Studio 1 is now flashing no it's not do a short flash on Chuck
lamp the Chuck lamp in Studio One is now flashing briefly no it's not okay so here's the state of voice assistance hosting it yourself it's okay like I think it's enough for me to make the move there are still some things that I think Alexa is very good at but it can't do AI yet which is pretty compelling for me to host my AI locally and have it talk to me on a voice assistant that I can deploy many of throughout my house that's powerful now you may be thinking what about a screen Alexa has
screens this doesn't have a screen hold on like I said you can do pretty much whatever Hardware you want to create a Wyoming satellite one of those is this this little guy here yes it does have a screen this is the ESP 32 S3 box it's created by Express if the people who make the esp32 and this is a specialized one that has a screen and it's designed to be a voice assistant now the S3 box is really cool screen's kind of small speaker kind of sucky but it's neat I might end up buying more
you can actually set timers similar to Alexa which this thing currently doesn't do so far yet yes it is missing some stuff from Alexa but I had so much fun doing this and I love that it's all local I don't even care I'll live with it and so all my family they'll have to cuz I'm going to replace Alexa now there's two things that I'm actually missing right now that I want to try and do it doesn't answer to Terry or he doesn't answer to Terry that teaching Terry his name is actually pretty easy I
got a link below in the documentation and home assistant walks us through creating our own wake word and what we're going to start is right here opening this link to a wick word training environment this is using Google collab you'll be using their data center to train a new word which is kind of funny when you think about it like for me I thought okay I'll just type in a new word and it will just recognize it but no you have to train an AI model to be able to recognize the sounds of Terry and
this will be fairly quick the first thing we'll do is enter our awake word right here now let's just try Terry and then click this play button on the left and it's done now we're going to scroll down a bit and click on this play icon down here and see if it sounds right carry carry carry what so you may want to do some fanatic spelling so I'll go back up here and change it to ter e let's try that let's see how it sounds Terry Terry better once you're happy with that go to the
top and click on runtime in the menu and then run all ready set go this might take a minute so get some coffee right now it's using machine learning to train a new wake word for your voice assistant when it does finish you should see two files downloaded automatically a TF light file and an onyx file this is what we'll upload to home assistant to give it a voice here in home assistant we'll need one add-on to make this work so we'll go to settings and and then click on add-ons and then we'll click on
the add-in store down here or add-on store and we'll search for Samba there it is what this will do is allow us to connect to home assistant's file system and upload files so we'll click on that and then click on install give it a second little coffee break and then before we start this we'll go to configuration at the top here just click on that and we'll set a password and then scroll down a little bit and click save then get back over to info and click on start now how you connect a home assistant
using SoMo will vary depending on the operating system you have for me I'm rocking window Windows I'll show you that real quick so first we'll take note of the IP address mine is 1921 16883 so from here I'm going to launch my Windows Explorer and now we'll have documentation below for the other operating systems and then right here at my I guess it's the address bar right at the address bar click in there and we'll type in back SLB slash our IP address back slash share it should pop up and say hey log in put
in the username home assistant and the password we just set and go awesome we're in now here we're going to create One Directory we'll right click and say new folder folder directory same kind of thing and we'll name it open wake word just like that we'll jump in there and then we'll drag our two files our TF light and our Onyx file right in there done now getting back to home assistant we'll go to settings Voice Assistant jump into our voice assistant and scroll down until we see open wake word we'll change it from OK
Nabu to there's a new one right there Terry can't help with that I just realized I said it it's so funny to Terry and update fingers crossed this should give Terry a new name Terry oh it already worked the living room room lights have been turned off would you like to turn on the lights in a specific area no Terry what year did The Matrix come out the Matrix film was released in 1999 however the movie's story line and Main release date is also subject to multiple variations as it includes different Matrix versions or chapters
Terry has a name also I don't like the voice it has I want my own custom voice to be Terry this Terry now I tried to make that happen in this video I couldn't figure it out I spent hours I spent a whole day just troubleshooting it and nights and I ignored everybody and I couldn't do it but I'm going to do it in the next video I don't care how much coffee it takes I'm going to do it
Copyright © 2024. Made with ♥ in London by YTScribe.com