No but you do need wyoming configured I believe.
1 Like
That totally depends on what you want. You can get everything running with Piper/Whisper, fully local, no expenses. And that works good to very good. The only thing to mention, is the hardware factor - you need something powerful, a Pi3 is nothing to work with.
But if you want the “cool” voices from the Azure Cloud, you need to get a subscription. It’s more like a plus-model, where you can use everything for free and local, but if you want more, you’ll have to pay. 
2 Likes
I was able to setup an app to detect a wake word and send the audio to a voice assistant pipeline.
3 Likes
Thanks dude for the reply
1 Like
Thanks for all the hard work! I am really looking forward to being able to (almost) completely switch over from using the Google speakers.
There are things that HA still won’t be able to do on it’s own (answer questions such as “what is this or that”, “where can I find X product” etc).
-
Do we have any idea yet as to what kind of hardware will be needed to use this with Wake word detection? I have no intention of actually buying them now but it would be helpful to be able to get at least a ballpark idea as to how much I’ll need to spend to make it happen. I don’t like to get hit with large unexpected expenses.
-
One thing I currently use my Google speakers for is to stream music – typically from the tunein app as they have my preferred local radio stations. When I set up my HA Voice satellites, will they be able to handle this task?
-
Once I have HA Voice satellites, I’d like to mute the microphones on all my Google speakers (so they won’t listen – improve privacy). If I make a request to HA Voice that needs to be handled by Google Assistant, can HA pass that request on to one of the Google speakers (preferably a specific one).
BTW… In my ideal world, I could use an old android phone and/Android tablet as satellites. I have a bunch of them lying around. This would give the added ability to run a dashboard on the same device. A phone that sits on my desk with the time shown and a few controls and it would also allow voice control.
1 Like
Help me to understand?
Can i translate intents to another language too? Right now Estonian language has only turn on/off intents.
Yes, you can always setup custom_sentences.
But don’t you think, it would be much cooler, to just work on the language pack for Eesti? So everybody, including yourself, can benefit from the translation work you’d do.
If you need help, let us know! 
I really want to try this! So My question is this. What device can I easily purchase, will be accepted by my wife (like an Amazon dot) plug in and will sit on my kitchen bench with a microphone and small speaker (for voice feedback and to use with HA notify announcements) to use with HA Voice? Where can I purchase one. Thank you
Read the first post of this thread. The ATOM Echo is cheap. It’s not a nice fancy Google / Amazon type device, those don’t exist for HA Voice yet.
1 Like
If you or your wife are expecting “Amazon” type performance for acceptance, you will be disappointed. We are not there yet, and most current focus I have seen and used is HA related. Not general smart speaker stuff. Maybe one day but not today.
Sure - but all I really want it for is to accept HA entity on/off triggers and run HA txt to audio notify responses back. I’m not interested in Amazon “type” performance, AI or playback of music etc. I’ve got better speakers for media streaming. Will this Atom Echo do the job of simple HA voice commands and notifications? Thanks
I have the echo, it does work for voice control it also works for notifications but it has a small speaker and can be hard to hear for notifications. If you are near the unit to speak you are close enough to hear responses or notifications. For now it is also push a button to talk. No wake word.
Does anyone has an example of how to open covers to a specified %?
Create a s riot that does it and expose the script to the local assistant. The great thing about scripts is you don’t have to say “turn on”, just the script name or alias so if you gave the script a voice alias of “fifty percent” you could bring up the assistant and just say “fifty percent” and the script would run.
I don’t think it’s a coincidence that express if just announced the Box 3 running an esp32-S3 with 2 microphones and a speaker. Heck, they even referenced Home Assistant and willow during the announcement.
The company also highlights the device’s compatibility with Home Assistant, described as an open-source and cost-effective alternative to Amazon Echo and Google Home. Further details about the Home Assistant can found on the Willow GitHub repository and its Wiki.
Specifications listed for the ESP32-S3-BOX-3 include:
Memory/Storage:
16MB Quad Flash
16MB Octal PSRAM
Display:
2.4” LCD display w/ capacitive touch
Audio:
2x Microphones
1x Speaker
Connectivity:
Expansion connector (PCIe x1)
Expansion:
1x 36-pin PCIe connector
USB:
USB input Serial/JTAG
Other Features:
1x Power LED
1x Mute LED
1x Mute button
1x Boot mode button
1x Reset button
Power:
5V (via USB Type-C)
Thank you for this feature and series of articles. Will Assist be able to be voice activated on google nest/xiaomi smart speaker? Instead of voice activation “hey google” will I use “ring assist”? Or should I not expect this option?
Welcome to the forum!
Unfortunately, you can expect this not happening. Simple reason is, HA has no way to change anything on Google Home/Alexa or whatever. These are closed source products, so what counts for them as “wake word” is hard coded into their firmware.
But I’d advise to take a closer look to all the possibilties that are already available for smart speakers in combination with HA/ESPHome. It starts with the very nice m5 Atom Echo and doesn’t end with the availability of a “standard” bluetooth speaker combined with an ESP32 or a Pi. Cheap, effective and works great with HA.
And to make things even more comlplicated, you can always go another route, where you use the “correct” wake word (“Hey Google”) and just leave the rest to HA. This is possible, but not to change the wake word.
Snagging some atom echos soon ahead of chapter 4 
Nobody stops you, but the Atom Echo is the nifty little thing, that was already introduced. The “new” device, where the “Year of voice” will work on now, is the ESP32 S3… I did not understand it fully in the last release video, as it was just mentioned in a side note from Paulus.
You might want to wait a bit, I suspect some new infos in the “birthday video” next week (or this week?).
Hi there, do you mean the esp box?
Hi everyone!
I greatly admire the developers behind this year’s Voice Project and Home Assistant (HA) Project as well. Thumbs up!
This chapter has captured my attention, and I have some questions regarding languages that are not commonly used. My native language is Slovenian, and from what I’ve learned so far, I should use Nabu Casa, which is very tempting. I would like to contribute to this community by paying for a subscription, especially since my home is becoming increasingly smart thanks to HA. However, I am particularly interested in using voice features in HA, but only if the Slovenian language is supported.
Do you have any information on how Slovenian (or any other less commonly used language) is integrated into this project compared to English?
Second question:
Is there any other esp32 (for use with Esphome) device as M5Stack for this? I was thinking someting like esp32 with mic and rgb light onboard to cover it as bulb and use it also for ambient light?