Rare to see product owners develop good products that would work without their company servers and API dependencies. Thank you for your dedication and development for the DIY communities.
Couple of things to make this an awesome product. Firstly USB C. Add temp, humidity, Co2, Tvoc and PM2.5 sensors if you can. Also if it supports Bluetooth presence system it will be an brilliant.
@@KinCony I built 5 of these, ruclips.net/video/YmqtMTO5NVc/видео.html one for each room. The senors are cheap and excellent as long as they are calibrated. So I kept the sensors outside once I built it and then set it to be 400 ppm for fresh air outside. Would love to replace those. An all in one.
I built my own "Jarvis smart speaker" (KinCony) about a year ago. If you use Faster-Whisper in a docker container that uses an old NVidia GPU, it will respond quickly like you are talking to Alexa or Google. The delay is not in the device he is demonstrating but in the services it is using. The other key is to setup OpenAI Conversation and Extended OpenAI conversation on your Home Assistant. This will OP your HA beyond Google or Alexa and for only pennies a month.
I've built a couple of voice assistants with S3s using the same microphone, and with the mic on the top of the unit they've not proven very good mics. With the sound of my PCs fans not far away it struggles to hear me reliably, yet the Echo in the next room does. Really needs USB-C though. Good job, best of luck with it.
USB-C is absolutely mandatory and non-negotiable. Everything else looks awesome. This is a MUST HAVE. Now in the "nice to have' graph is a bluetooth/zigbee hub variant along with ethernet port (maybe integrated Sonoff zigbee ?) This may be priced as a higher tier with more profit for you at the top. And the end variant , must be super luxurious painted in black and subtle gold instead of white and gray and have all the features above along everything you can put inside it - PM2.5 sensor, humidity and temperature, toc , co2 , presence detection etc. However I again USB-C is MUST HAVE in 2025.
Can you maybe update it to usb C. :) will be ordering today to test with my home assistant. Might have to buy a usb C port and try and mod it. 😊 Love your work
Did you test up to what distance the microphone works? And is the orientation of the device important? What I mean is, do I need to turn the microphone to point in the right direction or does it work from every direction.
i have tested in the room, distance 2-3 meter also is OK. microphone support every direction. actually the microphone on the back of BOX, i am not sit in front of microphone.
The reaction time is depending on the language model and the beam size, which you can choose in the configuration of whisper. Small is faster but less accurate. The speed is also depending on the compute power of your Pi/server that hosts Home Assistant. More power, faster reaction. If you use Home Assistant cloud you will be faster as well (but using a cloud service).
It would be interesting to see a solution not based on ESP-HOME and Home-assistant. Adding a second microphone and implementing echo cancellation. transfer of processed data from the microphone via UDP to the server. VAD (Silero or webrtc) + STT Whisper (Vosk ru) + TTS Silero(ru) analog en.
Way too slow to execute voice commands! My Google speakers execute commands so much faster through the cloud. I haven't seen a HA voice assistant yet worth switching over from Google/Amazon (nobody cares about Apple). Hope someone makes one in the near future, I would love to ditch Google in a Florida swamp.
@@KinCony Nope, have tried ESPHome devices on an Intel 12900K and they were all equally as slow. RPi 5 8GB no real difference. And when all the other major players in voice assistants can do this so easily over the cloud you have to ask yourself what the real problem is. There is still no competitive HA voice assistant worth investing in yet.
@@URackADisciprine Agreed. There are no solutions yet that can even be considered in the same ballpark as the big three when it comes to performance on so many levels. Running locally is much preferred but not at the cost of having to wait 10 seconds for every voice command to be executed.
Do the product demo at the start of the video. I'm not going to sit through 20 mins of the internals and how to configure it unless I can see that the user experience is good.
Rare to see product owners develop good products that would work without their company servers and API dependencies. Thank you for your dedication and development for the DIY communities.
thanks
Couple of things to make this an awesome product. Firstly USB C. Add temp, humidity, Co2, Tvoc and PM2.5 sensors if you can. Also if it supports Bluetooth presence system it will be an brilliant.
thanks for suggestion.
Will be expensive
@@sirmax91 CO2 sensor is expensive if the value is good.
@@KinCony I built 5 of these, ruclips.net/video/YmqtMTO5NVc/видео.html one for each room. The senors are cheap and excellent as long as they are calibrated. So I kept the sensors outside once I built it and then set it to be 400 ppm for fresh air outside. Would love to replace those. An all in one.
@@guylast9516 Mainly because the sensor is inside the plastic box, the air circulation is not very good
I built my own "Jarvis smart speaker" (KinCony) about a year ago. If you use Faster-Whisper in a docker container that uses an old NVidia GPU, it will respond quickly like you are talking to Alexa or Google. The delay is not in the device he is demonstrating but in the services it is using. The other key is to setup OpenAI Conversation and Extended OpenAI conversation on your Home Assistant. This will OP your HA beyond Google or Alexa and for only pennies a month.
thanks share your suggestion, yes, the server CPU is important for respond speed.
Any chance there’s a guide you followed that you can link to? I have an old nvidia GPU system I can use.
@@rrkumar78 ??
I've built a couple of voice assistants with S3s using the same microphone, and with the mic on the top of the unit they've not proven very good mics. With the sound of my PCs fans not far away it struggles to hear me reliably, yet the Echo in the next room does. Really needs USB-C though. Good job, best of luck with it.
thanks for suggestion.
USB-C is absolutely mandatory and non-negotiable. Everything else looks awesome. This is a MUST HAVE. Now in the "nice to have' graph is a bluetooth/zigbee hub variant along with ethernet port (maybe integrated Sonoff zigbee ?) This may be priced as a higher tier with more profit for you at the top. And the end variant , must be super luxurious painted in black and subtle gold instead of white and gray and have all the features above along everything you can put inside it - PM2.5 sensor, humidity and temperature, toc , co2 , presence detection etc.
However I again USB-C is MUST HAVE in 2025.
Thank you very much for your suggestion
Ordered Yesterday :-) looking forward to use it soon.
thanks
Can you maybe update it to usb C. :) will be ordering today to test with my home assistant. Might have to buy a usb C port and try and mod it. 😊 Love your work
because our plastic shell already made by USB-Mini. don't worry, the AS will include one USB-Mini Cable in package.
Awesome 👌
Nice product. But really an ancient mini usb 🤦. What is wrong with usb c. Would buy one if it had usb c.
thanks for suggestion.
Did you test up to what distance the microphone works? And is the orientation of the device important? What I mean is, do I need to turn the microphone to point in the right direction or does it work from every direction.
i have tested in the room, distance 2-3 meter also is OK. microphone support every direction. actually the microphone on the back of BOX, i am not sit in front of microphone.
Is there a way to reduce the waiting time between the speak input and the output action? I noticed Alexa thinks faster
three different option have different delay listening time. you can test
The reaction time is depending on the language model and the beam size, which you can choose in the configuration of whisper. Small is faster but less accurate. The speed is also depending on the compute power of your Pi/server that hosts Home Assistant. More power, faster reaction. If you use Home Assistant cloud you will be faster as well (but using a cloud service).
How can I buy your products in China? Do you have a store on Taobao?
sorry, not sell in china
@@KinCony It's a pity. And strange.
Does the product support vietnamese?
yes, supported, all supported language list, you can see here: developers.home-assistant.io/docs/voice/intent-recognition/supported-languages/
I think it's take a long time to get responses?
I think alexa speed than it?
it according to the wakeup word and server's power.
Very good.
Does it support Arabic language?
I hope that
yes, supported, all supported language list, you can see here: developers.home-assistant.io/docs/voice/intent-recognition/supported-languages/
What about other languages? Like Latvian?
yes, supported, all supported language list, you can see here: developers.home-assistant.io/docs/voice/intent-recognition/supported-languages/
It would be interesting to see a solution not based on ESP-HOME and Home-assistant. Adding a second microphone and implementing echo cancellation. transfer of processed data from the microphone via UDP to the server. VAD (Silero or webrtc) + STT Whisper (Vosk ru) + TTS Silero(ru) analog en.
locally will faster, if use server will have delay time i think.
Way too slow to execute voice commands! My Google speakers execute commands so much faster through the cloud. I haven't seen a HA voice assistant yet worth switching over from Google/Amazon (nobody cares about Apple). Hope someone makes one in the near future, I would love to ditch Google in a Florida swamp.
This mainly depends on the performance of home assistant server's CPU speed.
@@KinCony Nope, have tried ESPHome devices on an Intel 12900K and they were all equally as slow. RPi 5 8GB no real difference. And when all the other major players in voice assistants can do this so easily over the cloud you have to ask yourself what the real problem is.
There is still no competitive HA voice assistant worth investing in yet.
@@URackADisciprine Agreed. There are no solutions yet that can even be considered in the same ballpark as the big three when it comes to performance on so many levels. Running locally is much preferred but not at the cost of having to wait 10 seconds for every voice command to be executed.
Do the product demo at the start of the video. I'm not going to sit through 20 mins of the internals and how to configure it unless I can see that the user experience is good.
@@CountZilch thanks for suggestion, next time, let demo video at begin time.