This is really great! Thanks for the video. I wonder when someone will finally sell this device as a completed retail, working unit. For maybe $30 or so. This would be great for the HA community.
I have to give this a try... I placed my parts order. I hope you expand on the previous ESP32 Home Assistant too. Would be great to see features added to it. Thanks for the videos.
What I would like to see on this unit and the one you built before, is an audible indication that it is listening. The visual indication is only good if you are in line of sight. Sometimes these units do not hear and there is no way of knowing.
@@SmartHomeCirclethat would be awesome. My build is working really well, very responsive, based on your design. An audible "listening" tone would make this top notch.
The wake-up word is openwake, but can stt and tts support customization? Instead of using Wisper and Piper. Because I use funasr for stt and edge tts for tts
Wake word is instantaneous. But interpreting commands depends on the speech to text model you run as part of whisper and how powerful the machine is. In my case I am running an i5 6th gen processor.
Hm, why so quick reaction of voice assistant? Its amazing speed. In other demonstrations with the same setup of home assistant that a saw was a noticeable lag
I had created video a while ago, such that you can send the output of the voice assistant to a mediaplayer in home assistant. But it was for ESP32 and which supported the media player service. You can try it with the ESP32-S3, but I am not sure if that would work. Here is the video: th-cam.com/video/RIpnyTJmJJ8/w-d-xo.html
@@SmartHomeCircle I have used your code and flashed it. @formatBCE code has ok nabu but I dont see where it says jarvis in your code? Am I missing something? In the Video you didn't set the wake word in the voice assistant in HASS.
@@c0deless to have custom wake word at current stage you have to use streaming to HA, with Openwakeword on HA side. There's several wake words trained by Kevin for Microwakeword (which works on device), but no simple way to train your own. Also, only "okay Nabu" is pretty good, since it's using real voice samples. Others are trained on synthetic dataset, generated by TTS engines, so aren't very good at picking up accents and stuff.
This is really great! Thanks for the video. I wonder when someone will finally sell this device as a completed retail, working unit. For maybe $30 or so. This would be great for the HA community.
The board itself costs a bit more :) Then add speaker, enclosure and manufacturing - it will be more like 50-60$
Finally able to control the volume. That’s amazing. Thank u for the great guide.
I have to give this a try... I placed my parts order. I hope you expand on the previous ESP32 Home Assistant too. Would be great to see features added to it. Thanks for the videos.
@@tsmot911 i would be keeping the previous one also updated..
What I would like to see on this unit and the one you built before, is an audible indication that it is listening. The visual indication is only good if you are in line of sight. Sometimes these units do not hear and there is no way of knowing.
Ok. that is a good idea.
I might make a video around this.
@@SmartHomeCirclethat would be awesome. My build is working really well, very responsive, based on your design. An audible "listening" tone would make this top notch.
@drumslapper I am really glad it works well for you.
Enjoy! 🙂
Good Job 👍 Tank you for this Video
Amazing and impressive!
Great explaination..thanks
Glad it was helpful! 🙂
The wake-up word is openwake, but can stt and tts support customization? Instead of using Wisper and Piper. Because I use funasr for stt and edge tts for tts
Nice tutorial!
It would have not been possible without your help 🙂. I gave a small appreciation towards the work you did in the video. Thanks for this @formatBCE
@@SmartHomeCircle BTW i just posted 48kHz firmware there. Take a look, sound is amazing.
much appreciated
Great project! Recognition seems to be really fast 👍
Great video mate, I am testing another local voice device soon!
That sounds nice 🙂..
Great video thank you. Have I missed something, i dont get the "on device" option you show at 8:00?
The on device is the name of my voice assist pipeline.
You can see it here : th-cam.com/video/XjUeJh2Ok3o/w-d-xo.html
How does it handle background noise and distance? Most current esp32 existing solutions have been a bit underwelming in these aspects.
It seems better than just using the inmp441 microphones. I have been using it for a week now and it seems to be better for around 3 mts.
@@SmartHomeCircle How would you compare it to the usual echos and google minis in term of picking up the wake world and interpreting commands?
Wake word is instantaneous. But interpreting commands depends on the speech to text model you run as part of whisper and how powerful the machine is. In my case I am running an i5 6th gen processor.
This is awesome thank you! Is there a way to do presence detection with Bluetooth beacons with this board?
You might be able to but I am not sure of that. You will have to try it out. . Currently I have only shown the voice assistant part.
Is it possible to use this with a Sonos one sl? Just it doesn’t have a 3.5mm jack
Apparently not. Since it s a wireless speaker.
I wish you didn’t edit out the gaps so viewers could get a true sense of how fast/slow it is.
@@ZiRo815 slow or fast reponse times of the command execution depends on where you running whisper and piper. This device is independent of that.
Hm, why so quick reaction of voice assistant? Its amazing speed. In other demonstrations with the same setup of home assistant that a saw was a noticeable lag
@@ruslanUsifov there is a small lag of about 3-4 secs.. which I have edited out of the video..
Would it be possible to use a Sonos speaker via wifi instead of a speaker connected to the 3.5mm jack?
@@fugixi you can use it as speakers in Home Assistant via wifi.. but not as voice assistant speakers.
There is a hack, that allows routing TTS to other players in Home Assistant. But it's really buggy and hacky way to run things.
@@formatBCE Yeah, I can imagine. Thanks.
I had created video a while ago, such that you can send the output of the voice assistant to a mediaplayer in home assistant. But it was for ESP32 and which supported the media player service.
You can try it with the ESP32-S3, but I am not sure if that would work.
Here is the video: th-cam.com/video/RIpnyTJmJJ8/w-d-xo.html
Can you change the on device wakeword or provide different wake word files?
Yes you can.. but you will have to train the wake word file by yourself.
I think you said Jarvis in this video. How did you change that?
@@guylast9516 you have to set the wake word. Check my article in the video description.
@@SmartHomeCircle I have used your code and flashed it. @formatBCE code has ok nabu but I dont see where it says jarvis in your code? Am I missing something? In the Video you didn't set the wake word in the voice assistant in HASS.
Check the micro_wake_word section in the yaml code, Its in the models property.
Is the language supported in Korean?
That is based on if Home Assistant support korean language.
Oh no I heard the words HomeAssistant.... no, just no, its a migraine that software. and I've heard its true that it grows back
Sure seems like a lot of steps.
Is this device powerful enough to have a custom wake word?
It has ESP32-S3, so it would be. Currently the process for creating custom wake words still requires a power machine and the process is tedious.
@@c0deless to have custom wake word at current stage you have to use streaming to HA, with Openwakeword on HA side. There's several wake words trained by Kevin for Microwakeword (which works on device), but no simple way to train your own.
Also, only "okay Nabu" is pretty good, since it's using real voice samples. Others are trained on synthetic dataset, generated by TTS engines, so aren't very good at picking up accents and stuff.
You've got a face for radio.