I followed your tutorial, but I can't get the snapclient to output sound. I tried setting the output device, but it can't seem to find the mic hat. Also, during the tutorial I was able to execute the paplay command, but after completing the steps it errors with: Failed to drain stream: Timeout. Could you help me with that?

Unfortunately, after second try to do this i have to admit, that pulseaudio makes delay between wake word detection and voice capturing totally huge, which renders satellite unusable. I'm on Pi-Zero_2W, and it's awful. I really wanted it to work... Also for some reason pamixer isn't available for my OS (Lite, legacy)... Will ditch this, since i already have good working setups on ESP32-S3. I really wanted it to work...

Hello, quick question. Is it possible to have more than one wake word for the voice assistant, like different profiles. So for example I would have one that’s “Hey X”, my partner could use “Hey Y” and they would have different voices for each one. TIA

Here is a question for you how could I get it to announce messages for example tell me once it locks the front door from automation not from me asking if the door is lock where I can lock the front door with an automation and then after it locks said front door it notifies you through satellites stating that it just locked the front door I can get it to send me a text message

how can you control the leds they are bright everything else perfect thank you

What are the risks of exposing your enitties to OpenAi? I mean -e.g., safety and privacy risks, costs, etc.

6:50 Black Cherry? No way. I vote Raspberry.

would really love tha unraid llm setup vid. i have llama.cpp running on my unraid with the functionary model installed but responses take forever in the llamma.cpp web ui. what am i doing wrong ?

This has been so helpful! Thanks! Where do we go to figure out how to evaluate the quality of the mic vs. the auto gain, noise suppression, and other tweeks? Im having trouble getting it to recognize aliases in home assistant. Im using a logitech USB mic and it recognizes the wake word from across the room but only recognizes the alias when Im 5 ft or less from the mic.

Thank you, I built the room-assistant version and its amazing. It takes a min to update but thats good enough to adjust the thermostat. Now i need to figure out more about presence detection and hysteresis so that things don't toggle the instant I'm out of range or if there is a blip in the measurement for a moment. What would it take to update room-assistant for the modern era?

I've been looking at doing something similar to this, although my setup is perhaps slightly more 'traditional'. I have a giant server with the potential for a full size modern GPU, and I'd really only need a bunch of mic remotes for each room. While I do want GPT functionality, the main purpose is to actually continually record audio, and only save it if I tell it. It's a way for me to hash out ideas/solve problems/etc, when I don't have pen and paper in front of me. I can then have an LLM ingest the audio, run a summariser, and log it into my notes database along with the audio.

Any update on this? It looks great!

i have gone through this step by step 5 times and I still cannot get snap ○ snapclient.service - Snapcast client to run. Any help would be appreciated.

what actual speakers are you using? I've gone too far down the rabbit hole trying to figure speakers and amps out

So if you throw a dedicated powerful gaming pc at it with strong cpu and gpu, do u think you could get a response time to that of alexa? Just wondering..or would it still have a delay..

I'm having an issue with it connecting to my home internet. Spectrum makes us change the settings to enable 2.4 gig. I do that and its not picking up. Also have wiped the SD card and retried several times. Any ideas on what I might be missing? I'm sure it has to do with changing my WIFI over to 2.4. Thoughts?

How can I modify the verbose answers from ChatGPT to something shorter (like "ok" or a "beep")? I tried various instructions in the settings to no avail... :(

This is great, I am actually using it as a work-around for Node-Red flows to provide custom tts.speak responses back to the satellite, which isn't currently supported by Node-Red BETA "Assist" Sentence node. However, two issues I want to tell you about: 1) I found that after installing pulseaudio it drastically reduced the overall volume and max volume ability of my satellite. To resolve, I had to edit the "/etc/pulse/daemon.conf" file and change Flat-volume = yes. Which allows the max of the audio hardware, not software limited. Reboot required after change. 2) My satellite is connected to a USB Conference Speaker/Mic... this is for better audio and voice capture. I mapped the volume buttons on the conference speaker to change the Satellite Audio. Again installing pulseaudio altered the system and changed the amixer device reference from 'PCM' to 'Master'. This means I had to change my mapping to listen for USB events for volume up/down and alter 'Master" audio device. I also had to add a reference in the snapclient.service file to use --mixer hardware:Master. Notes for others who may experience the same. I'm using this conference speaker: www.amazon.com/dp/B0B4JTMQ9H I have the volume keys mapped, which allows me to locally control speaker volume and allows me to mute the mic from the device, without going on Home Assistant.

Are you on a paid plan with OpenAI? How high is your usual monthly fee?

Is it possible to clone the micro sd card to another pi? So you can have two satellites

I am actually waiting for a good and easy solution to bring in a home-assist that is like alexa etc. but has the knowledge like ChatGPT. I hope one day there will be a device for it. I would probably pay 100-200 for a device like that excluding the monthly payment which I'm also willing to pay to get a good and usable AI Home assisted. I also hope the text-to-speech will improve like ChatGTP 4o has atm. which is pretty good. I'm a basic user, I'm not in the game like most of the people watching this, but I would love to be able to control all my smart lights and not have a stupid ai that knows nothing. All the AI assistants like Google, Siri, and Alexa are just not coming along to today's ChatGPT.

Super good video. How do you send TTS to a Satellite?

Why do you use a slow pi?

I'm not sure why, but I had hundreds (400+) of temperature and humidity sensor entities which exceeded the API token limit. Lots of cleanup and now things seem to be working.

I just love the work that you are doing! Tried to go into the exact same direction but you are far more advanced. I also would love to see this setup as a product, because I realized that progress on HA also impacts my custom setup and therefore needs a to of attention and maintenance. So yes, I absolutely would be interested if you can offer this as a product in the future.

Just followed this method to a 'T', and still have no sound output from the device when playing music, or using the voice wakeword command. Its listening, because when using the wakeword, the LED's light up, and the command is executed. However snapclient/snapcast music playback does not work, and the sound output does not work. The error i get is Jul 07 09:59:04 wyoming pulseaudio[510]: Failed to open cookie file '/var/run/pulse/.pulse-cookie': No such file or directory Jul 07 09:59:04 wyoming pulseaudio[510]: Failed to load authentication key '/var/run/pulse/.pulse-cookie': No such file or directory Jul 07 09:59:04 wyoming systemd[1]: Started pulseaudio.service - PulseAudio system server. Jul 07 09:59:07 wyoming pulseaudio[510]: Failed to find a working profile. Jul 07 09:59:08 wyoming pulseaudio[510]: Failed to load module "module-alsa-card" (argument: "device_id="0" name="platform-3f902000.hdmi" card_name="alsa_card.platform-3f902000.hdmi" namereg_fail=false tsched=yes fixed_latency_range=no > Jul 07 09:59:14 wyoming pulseaudio[510]: Failed to find a working profile. Jul 07 09:59:14 wyoming pulseaudio[510]: Failed to load module "module-alsa-card" (argument: "device_id="0" name="platform-3f902000.hdmi" card_name="alsa_card.platform-3f902000.hdmi" namereg_fail=false tsched=yes fixed_latency_range=no > Jul 07 09:59:14 wyoming pulseaudio[510]: ALSA woke us up to write new data to the device, but there was actually nothing to write. Jul 07 09:59:14 wyoming pulseaudio[510]: Most likely this is a bug in the ALSA driver 'snd_bcm2835'. Please report this issue to the ALSA developers. Jul 07 09:59:14 wyoming pulseaudio[510]: We were woken up with POLLOUT set -- however a subsequent snd_pcm_avail() returned 0 or another value < min_avail. But when using the command paplay /usr/share/sounds/alsa/Front_Center.wav the sound outputs fine. I ran the command pactl set-default-sink alsa_output.platform-soc_sound.stereo-fallback to set the output to the 2mic pi hat. Any ideas

eagerly awaiting an update.

Amazing stuff, man! You are talented and gifted with a good brain , I’m gonna stick around here … I’ve build a full size droid … and trying to figure out how to stick an AI assistant inside him … this units you installing , could be the answer ( if you like to check him out , jump on my channel and look for AI droid video I recently uploaded ) Looking forward to your stealth ai responding installation..

How slow is slowGPT ? I'd prefer to run that, but not it it's toooo slow

Can I suggest you add {{ area_name(entity.entity_id) }} to the Custom prompt so that Open AI understands which devices are in which area. If people use areas as rooms it makes it more useful. This allows "What is the temperature in <area>?"

for some reason, the wakeword in home assistant takes lile 25 seconds.. sometimes more up to 500sec.. any idea?

Really super cool video! It's a pleasure to listen and view it. Great job!!

You need to mention that local llms are only slow if you're Running it without a video card. As soon as you add a GPU the llm is fast. A gtx 3090 is 💋🤌

Memes per second were reaching cringe levels my guy

Can we connect multiple Voice Satellite device to one Home Assistant? Because we have multiple room in the house.

I had bought an new Intel ARC A770 graphics card with 16GB RAM for running LLM. It was relatively cheap (under 300 Euros in germany). And it should outperform an Nvidia RTX 4060. I haven't tested it yet since it was delivered yesterday and now i am setting up an new server (Proxmox on an AMD Ryzen 7 5700G). I can't wait to see more videos from you for building my own offline alexa system with interaction in home assistant.

Great video. I am just starting my journey in HA. Goal is to replace my Amazon Echos with this type of solution. I have watched all the videos and am sold on this path. Any update on when a localai video will be coming? Thanks for all you are doing to make this happen!

Very deep and informative, great video. Does anyone has experience streaming audio from tv to the multiroom, does it get crazy latency?

the future is here! great video

I'd recommend to you, to show the results in the beginning of the video, and then how to do it. :)

You have few link for each one? What is the difference?

The issue i am having with setting up ceiling speakers is actually cutting holes in the ceiling. :) we have a 1 story house and adding all those holes into the attic is not good..Which I had a conditioned attic and then i would not have an issue. Any suggestions of sealing ceiling speakers?

Question.. how is the pi powered? Where is the power coming from?

I get a very fast and high pitched TTS response using this setup with a small speaker connected via JST. I don’t know which setting affects the speed of the TTS response. Any idea?

Is it possible to ask chatgpt questions not related to home assistant ?

Hello friends, My problem is that if I follow this tutorial one-on-one, after installing the “Local Wake Word Detection” section, the RasPi no longer hears me or perceives me or whatever, but just no longer responds. [ Short info, there was always an error in the script during installation - you have to remove the "nightly" in the "requirements.txt" - I already found out thanks to the comments here ] And then when I check the status I get the error message. If I skip this part, everything works perfectly and the HA responds and carries out my requests. But just so that HA is listening the whole time and not the satellite as desired. Does anyone have a tip or suggested solution who has already recreated the whole thing and been able to fix this error? I would be extremely grateful to you!!! SOLVED!!! Thanks to a FB user in the Raspberry Pi group, my problem/error was solved. Thanks again to Daazed McFarland on this path and also thanks to everyone who also thought about my problem.

I got all my services correctly setup and checked its running with no errors, but when I use the wake word, it does not trigger on Home Assistant. what could I have done wrong?

Super interesting project! Just build my first voice assistant box based on a esp32 s3, an addressable led strip, little speaker and an inmp441 microphone for around 20€ and now im hooked 😁 it works beautifully with local wakeword detection. Not yet sure about the microphone performance in a loud environment but on the first look it works like a charm (with homeassistant cloud and chatgpt as the brain) :) maybe worth a look if you want to get up a "HomeX lite" with more competitive pricing! Would love a off the shelf version for a reasonable price 😊

Sorry fot the OT... what remote is that please ? :-) Thanks for the video

Is that (rasp+mic) sensible like alexa/ghome assistant please ? ... I mean using the wake-word. Thanks for the video.