Brilliant! I won’t cancel my ESP33 S3 box order after all!
I used the new .yaml for my S3-Box-3 but micro_wake_word is not working. Looking at esp32-s3-box-3.yaml there is no micro_work_word mentioned anywhere in the file. Flashing the ready-made voice project for this box works.
You are not using the correct yaml.
Use this one
2 Likes
Reading through this thread it seems that nobody could successfully run ESPHome media_player in parallel with voice assistant. Is this correct?
That’s correct, but a fix is already in the making so hopefully won’t be too far away
4 Likes
Thank you, that’s good news.
@starob, the correct one should be:
packages:
- esphome.voice-assistant: github://esphome/firmware/wake-word-voice-assistant/esp32-s3-box-3.yaml
1 Like
I already complained above that the volume of the box is very low. I installed the stock demo firmware again to check the max volume. It allows you to play a few MP3 files. Increasing the volume makes the playback pretty loud. Much louder than with ESPHome. This means its not a hardware issue but a ESPHome issue.
I opened an issue for it here https://github.com/esphome/firmware/issues/139
some time ago but no progress so far.
Does somebody know more about this problem?
2 Likes
I’ve noticed that if an intent does not have a speech response, then the Box 3 will get stuck on the thinking image. For example, saying the built-in intent “nevermind”. It also does this on some custom intents I have written that do not have speech response. I’ve waited 20 minutes to see if the Box 3 will recover, but it just stays on the thinking image and does not react to the wake word. The “Assist in progress” sensor also stays On. After I reboot the Box 3, it does work again, but the “Assist in progress” stays On until I speak another command.
I don’t know if this is something wrong with my setup or a general problem. Is anyone else seeing this? If so, I will write an Issue.
1 Like
I can confirm this sound behavior of the BOX-3. It’s a big difference. A fix would be great.
Getting this to the point where it can be used as a media player for notifications while also being a voice assistant would be awesome. Looks like they are working on that, so I hope they achieve this quickly.
It would also be nice to be able to utilize the sensors in the base. I would love that capability.
Hi Jason,
Radar and battery sensors are available. Add this to your device yaml:
packages:
...
esphome.voice-assistant-sensor: github://jaymunro/esphome_firmware/esp32-box-docks/box3-sensor-for-voice-assistant.yaml@sensor_dock
And if you are interested in exciting upcoming features I have been working on (continued conversation and display text):
Change this
packages:
esphome.voice-assistant: github://esphome/firmware/wake-word-voice-assistant/esp32-s3-box-3.yaml
...
to this:
packages:
# esphome.voice-assistant: github://esphome/firmware/wake-word-voice-assistant/esp32-s3-box-3.yaml
esphome.voice-assistant: github://jaymunro/esphome_firmware/wake-word-voice-assistant/esp32-s3-box-3.yaml@continued-conversation
...
I have a PR in draft that I am looking for testers on so it can be released on the esphome main repository. I’ll add a link to the pr with more description and video once I’ve got the video done.
1 Like
Thanks! I’ll test these out.
Ok, I tried these and could not get them to work. Here is the config file along with the logs from failed installs. The names of the logs tell you which of the packages were uncommented. Hopefully you can tell me what I did wrong.
https://drive.google.com/drive/folders/1xRoJ3QdyhAM6F1htR1cNLpeQ2c7Kfxjp
I had a look at your logs. The errors/warnings are typical of a need to do a clean build.
Not sure which you did first, but you need to do a “Clean build files” when installing new components (which will be many the first time round). Did you try that? It’s just a few items below “Install”.
This is common with the espidf framework.
1 Like
Ah, thanks! I was hitting the install button in the editor. Had to back out of the editor to find the “clean build files” button. It’s working now. I really like having the sensors and the last command/response saved.
Here is a demo video of this beta of Voice assistant with Continued conversation and text.
I’m looking for testers who can give feedback on issues, further suggestions for improvement, etc. I’ve tested it with the HA streaming wake word but it works much better with the on-device wake word. Once it is verified, I have a PR ready to go in ESPHome repository. Please leave issues and suggestions in the related issue tracker (not here because they will get lost) Feature: Continued conversation with text on screen · Issue #174 · esphome/firmware · GitHub
To install, in your device’s yaml change this:
packages:
esphome.voice-assistant: github://esphome/firmware/wake-word-voice-assistant/esp32-s3-box-3.yaml
...
to this:
packages:
# esphome.voice-assistant: github://esphome/firmware/wake-word-voice-assistant/esp32-s3-box-3.yaml
esphome.voice-assistant: github://jaymunro/esphome_firmware/wake-word-voice-assistant/esp32-s3-box-3.yaml@continued-conversation
...
If you change the wake word model, add another substitution for the wake word text as follows:
substitutions:
...
micro_wake_word_model: hey_jarvis
wake_word_text: "Hey Jarvis"
Note this is a beta and the repo above is temporary until after the PR is accepted.
Before doing an Install, be sure to do a Clean build files from the 3-dot hamburger menu in ESPHome.
2 Likes
Great @HJM!!! Just installed your code! Thanks for your work and effort! The best is continued conversation!!
The output sound is from time to time laggy, but ondevice wake word is much faster than on HA 
The date and time - how to configure the time and date? Mine shows Mon, 5 Jan, 1970 - 8:41 PM (both are wrong
The display text is working pretty good - except localization (it’s in EN) - is there any chance to get it working in system default language (mine is Slovenian for example)?
But I’m not sure about the fonts - it doesn’t show out characters (I think it’s ISO/IEC 8859-2).
Wake word: from time to time the I get in the log:
09:31:48][D][voice_assistant:349]: Speaker buffer full, trying again next loop
and after few minutes it’s ok and the box starts to listen to wake word.
I’m really looking forward on this (yours) code, I hope the PR is getting approval soon Thanks again for your work!!!
I finally succeed to install it, If you don’t have more than 4GB it will not compile, I update to have 6GB of ram and It worked, (didn’t use more than 4,5GB).
Right now, everythings seems to work, I gonna continue testing
1 Like
a few things I realized :
- after flashing it never trigger again… even after a reset but a boot cycle make it work again.
- the text is only in english would be nicer to be in substitions to be able to change it
- the text doesn’t support some char, I’m french and turn off is “Éteins” It show a black rectangle instead of the É
- would it be possible to have change the format of the date ? and time ?
I really like the time display in the center.
and the text also it’s nice too
maybe it could be nice to hide the say “okay nabu” which couldl be wrong since it can’t follow the actual wake word name but keep the rest of the text ?