For ESP32-S3-BOX, have you been following the Willow project? Looks really promising. The aim is to provide voice hardware for various projects, including Home Assistant Voice Assist.
I bought one, but right now the install / build process appears daunting. I’m hoping the promised simplified install process arrives fairly soon.
That device, its design, its fit & finish, & its capabilities (like dual mic) all look very promising.
Wow, I completely missed that! Will tinker right away! Thank you for this.
Great I now can pick dutch or English by dialing the right number.
Getting some AI to pickup my blacklisted numbers of my asterisk system is tempting.
Question.
My asterisk setup uses the google text to speech cloud system. I want to replace this with my local piper setup.
How can i directly send text and get sound from piper?
Is it possible to confirm that Piper will not replace the Nabu Casa Cloud TTS in the future?
It’s not stated clearly on Cloud TTS page but it seems it’s based on Azure Speech Service as we can see in the code).
The quality of the voices from Azure (or whatever NC is using) is way better in French than current voices of Piper. And even if I believe it can be improved a lot with training, I would still prefer to keep the option to use either Azure voices from Nabu Casa subscription over Piper.
Extremely unlikely.
There is something very strange about assist in the Hebrew language
When I try to use TTS in Hebrew, it only gives me the option of Home Assistant Cloud
The google translate option exists, but it cannot be clicked, (in general, I always use google translate properly in the Hebrew language)
I also have the same problem with the SST option in the Hebrew language, only the Home Assistant Cloud option appears, while the other options (Google Cloud Speech-To-Text) are found but cannot be selected,
what could be the problem?
Do you guys know if “assist” can be called in a “card”? Is there an “assist card”?
Not sure I understand the use case? You can just click the icon top right?
as an example , search has a seperate card available that can be “called” everywhere, even as a pop-up. Assist can only be "called"when clicking on the assist icon in the menu bar. However, I have created a custom menu bar in my dashboard. I want to click a (defined by me) button, providing the Conversation pop-up. Is there another way of calling Assist, so not by pressing the conversation icon?
I wish. Its completely unusable when you have kiosk mode enabled. Having it as a card or a service that can be called to activate the popup would fix this.
Late reply, but I am thinking of something like the Shelly Blu, but would also need a mic and speaker. Blu already has BLE ibeacon support, is a nice size/form factor, and uses a battery. A Blu with mic and speaker would be perfect for this and wouldn’t really take any more battery power since the mic would only be active when the button is pressed. If they made one, I would definitely buy and carry/wear.
https://www.shelly.cloud/en/products/shop/shelly-blu-button1
Edit: It looks like it already has a speaker so you can “find my device,” so it just needs a mic.
Edit2: It is Bluetooth only, so probably can’t be used for direct smart control. But, it should be possible to have a smart mic/speaker device and use the Blu’s button to activate listening on the smart mic/speaker. It would require some automation and maybe dev work, but I could see hitting the button on an iBeacon (like the Blu), getting a chime/chirp from a smart mic/speaker and then saying your command. That would be a nicer experience than going to and then picking up a device with a wire to hit a button on it.
Same here, did you manage to hide the assist icon somehow?
Also can’t remove the default configured voice… looks like this option isn’t very polished yet…
Nope, I did not 
same with 2023.6.x. I still don’t know how to remove it.
With card_mod themes. Have a look in the corresponding thread and/or ask there. I think that Marius already did it.
Thanks, will take a look there!
Don’t know if the below project would help. Someone managed to get wake word detection working using tensorflow lite on an esp32 and it seems to work pretty well. This would be great if we can use the m5 Echo
Not sure what project you talk about, but according to Blakadder, this looks promising as well
I recently went through this, I eventually got GPU working with Docker/wyoming-whisper. Details here: Home Assistant - Enabling CUDA GPU support for Wyoming Whisper Docker container | tarball.ca