#Atom Echo Ever Fixed?
1 messages · Page 1 of 1 (latest)
The echo firmware is working fine. it looks from your screenshot that either there is an incorrect firmware installed or the device needs to be re-added to the HA integration. The best steps to take would be.
- Delete the device from the ESPHome HA integration and also any duplicates
- Remove the the Device card in the ESPHome Dashboard and any duplicates
- Check that you are running the latest version of ESPHome - it will tell you this at the bottom of the dashboard. The current version is 2024.9.1
- Plug the device in to a computer that you can access https://esphome.io/projects/index.html from
- click
connectand select the correct com port for the device. If more than 1 port is listed unplug the M5 stack Echo and the port should disappear from the list, plug it back in and you should be able to identify which is the correct port. - click to install the firmware and follow the steps to add to wifi after the installation finishes.
- add the device to the home assistant ESPHome integration. The device should be auto detected and show at the top of the 'devices and services' page if not then add the device manually by clicking 'add integration' then 'ESPHome' at this point you will be asked for 'host' for this the easiest way is to use the devices ip address (you should be able to obtain this from your router)
- Once the device is added to the integration, open the ESPHome add-on 'Dashboard' and there should be a device card to 'adopt' - continue to adopt and install - to avoid any issues at this point don't change any of the preset values eg. name etc.
- Once the installation is complete you should be able to use the device.
You should have an option in the device integration page to select wakeword engine location. By default the 'on device' wakeword is set to 'okay nabu'
If at any point you have an issue with any of the steps above, please drop a message in this post and we can help to get you up and running.
I hope the above helps 🤞
You lost me at step 1. Go to settings, then devices, then m5stack atom echo, then do what to delete it?
Go to integrations > ESPHome and click
> Delete for your M5Stack ESPHome config entry
Ok. Got to step 7. After installing firmware, device being auto-detected, I get this
Esphome dashboard shows no devices, so I can’t look it up. Skip and reinstall firmware via esp dashboard? That seems to be what others have to do
It's weird that a device flashed with the ESPHome web installer requires an encryption key. It shouldn't have one. Are you sure you've flashed it properly? If so, does simply clicking "submit" resolve the issue?
Will start from scratch with screenshots. Had never worked any other way to the best of my knowledge. Figured the instructions were just obsolete
Second time with same steps worked. Now, how do I select my wakeword without things breaking again?
there should be a select entity in the device page
or do you mean selecting a model other than ok nabu?
Nevermind. I killed it. At step 8. After adopting, it installed new firmware and the device is bricked.
Was actually working until I tried step 8
After step 8, it looks like HA installs some firmware that bricks it. The device logs show a boot loop. Looks like the firmware generaged by the esphome plugin is bad
Been waiting over a year for them to get it working. Should I go for an alternative device?
There's bunch of people using Atom. If yours isn't working, it doesn't mean device is bad...
Which ESPHome version do you use? Did you update?
2024.9.1 - I keep updating each month in hopes that the next update will make the atom work. The boot loop messages seem to indicate that the firmware image generated by HA is too big to fit on the device, but it partially installs it anyways
Do most people just skip step 8?
I guess that would work as long as you’re okay with all the defaults and don’t want to change anything
Do I need to edit my config file as described in the solution here? https://github.com/esphome/firmware/issues/227
Any other suggestions, since your steps fail at step 8 when the integration makes a firmware image that is too big to fit onto the device?
I will take a look at this, test on my device and update you shortly.
In the meantime...
Did you install the firmware wirelessly or via serial/usb ? if wirelessly could you try via USB . by choosing the 'manual' option after clicking 'install'
This will then create a .bin file and you will be given the option to save the file at the end of compile. If prompted select 'factory' for firmware type.
Once the file is created and saved , visit https://web.esphome.io you will be able to connect to the device and install the saved file.
- UPDATE to test.
I have just followed the steps above with my test device and everything happened as expected and the install from the dashboard worked without issue.
I’m good until step 8, where it generates a firmware image that is too large to fit. The thread I posed a link to says that this is the behavior others see. Do you have a different version of the atom echo with more storage, or is there see trick to tweaking the firmware image that HA creates?
After you adopt and tap install, it compiles code for a few minutes and overwrites the firmware, right?
yes, after adopting, it then should compile and then install. There is only one version of the device and the one I have was purchased 12months ago, they all have the same flash size.
If the install isn't working as it should.
If you then try by clicking install again from the device card in the dashboard and following the steps https://discordapp.com/channels/330944238910963714/1287619920594468885/1287939274003451967
to install via usb. If for some reason the partitions on the device have become corrupt or some other reason Then flashing via usb 'should' rectify this
Ok. Starting over from scratch. Delete the item in HA called ESPHome web too?
yes you can delete that one, and just click 'install' on the other
and then select ' manual' not wireless
from the screen you pasted above
After deleting everything, I have this
dont click install for first use - that will create another ESPHome web device and will wipe the device again
ok if you have the time to go throught it now.. i am around for the next hours or so .. so will make it easier to solve in realtime 🙂
ignore that card totally
Ok, do to what next?
ok as you have deleted the M5 card we will need to start from the begining , but first you also need to delete the device and any duplicates from the HA integration which is in HA >>settings>>devices and services>>esphome
click on the 3 dot menu next to the device to then delete
and then when you get to the new adopt card popping up in the dashboard we can then go through that part step by step
ok ignore that .... that is because the device has now been flashed with the new device config ... but we will sort that . if you now go to the esphome projects page and install, add to wifi and add to HA as described in steps 1 - 7
that part is all working as it should... then we can take over and sort the final steps step by step ... so once the adopt card for M5 appears... let me know
yup connect
ok 👍
ok ... next and then add wifi creds
ok
yes.. and go ahead
click adopt on the M5stack card
at the end of this it will possibly fail again .. and then we can sort it manually. so just let it run its course
yes, it will do 🙂 the yellow warning are normal
looks good so far
Got this far every other time
ok
ok if you click 'stop'
bottom option
ok ... now go to https://web.esphome.io
yes ... and then 'install' and NOT prepare for first time
Choose newly saved file?
ok now go back to the dasboard and see if you can view the logs from the device card
you can still ignore the web one, that will dissapear eventually
ok , is the M% on the same vlan/subnet as your HA server ?
Yes
if you go back to https://web.esphome.io and click logs
at the bottom click 'reset device'
does that stay like that or is it looping
Just stays
If I install from HA instead, it does the looping
The thread I posted says that the image is too big to fit & has been a known issue unresolved for months
Hence my question if it’s gonna get fixed
can you go to HA>>settings>>add-ons >>esphome and click 'configuration' at the top and then 'use ping for status' and save
a fix was merged back in july for that particular issue. This is is the first time I have come across the problem since on here or the ESPHome discord ... so it may be something within your esphome installation causing it. as mentioned above it installed fine for me when testing earlier
i'll just check file
that all looks correct
thats all fine .... what shows in the device page in the integration?
file size is fine
ok so HA isnt seeing the device... are you able to find the ip address of the M5Stack from your router ?
It’s not on the network after bricking. Viewing logs via usb shows it hung after boot just after the ha-generated firmware gets installed. If I install it from HA via usb, then I get the boot loop described in the thread.
It does this, then hangs
Was unstable, but at least booted until recent HA updates
just looking at the one i have here just to compare logs
my logs are identical, you need to check your router
Check it for what? Isn’t this all local traffic?
to see if the device is showing connected to wifi
from the logs you posted above the device is not bootlooping
that is all the logs show is the bootloader and that all appears ok . so would indicate it is all ok otherwise it would loop
Flashing with an older firmware that was generated a month ago…
Ok. That was the one that made it boot loop. No dice
Is there a higher-powered esp32 device that will work as a voice assistant microphone? I’m using a Sonos for the replies, so I just need a mic that works
in the esphome dashboard - can you click 'secrets' in the top right and check that the credentials are correct
Looks right. Will retype and try again
not currently the only option is the ESP32 - S3 - Box - 3 - but that has a screen mic and speaker. The other alternative is a diy ESP32S3 and a mic and make your own
May have bern a typo in password. Re-flashing… a wrong wifi password would cause it to stop running and hang after boot with no logs?
yes as it needs to be connected to wifi to retrieve the ESPHome logs
I thought when connected usb we were looking at real-time logs from the device
you are ...however the firmware doesn't appear to be configured to produce serial logs after the bootloader... I am just looking into that now
i'm also just making a full config to go into the dashboard rather than the scaled down version using packages
Password issue…. Getting further…
ok 🤞
are you using on device wake word or openwakeword?
Looking. Don’t know the difference
Only option I have for “wake word engine” is openwakeword
On-device = Alexa or special hardware?
on device is built in to the M5 and you can use either of these wakewords Alexa, ok nabu, hey mycroft or hey jarvis. THese need to be configured on the device. or openwakeword is setup in HA from the assist page
Ok. Open wakeword
ok so you don't need to do anything else with the device and flashing
that should all now be configured and working
Ok. Is still responding to the original wakeword, not t
He newly selected one
Restart device or HA?
can you show me the integration page
ok it is set to on device
Change this?
ok so change tfrom on device to 'home assistant'
then whatever wakeword is selected in the page you have just shown should be the one that works
Okay. Then flash? Reboot something?
great ... we got there in the end 🙂
The problem I have had was that the device stops responding to the wakeword after a few uses. There was some known issue with the ring buffer
Sometimes you could get it to work by shouting loudly at a close distance. Other times, nothing at all. Rebooting the device made it work for a fe
W more times or hours
Yep. It worked about 3 times and now only works if you shout the wakeword loudly
There is constantly work being done to improve the experience.. it is still a relatively new thing and with people reporting issues etc it will just improve with time. But it relies on people contributing and providing feedback in order to make improvements. On device wake word will be the better option in the long run and voice samples are being collected by volunteers to improve recognition. Due to the complexity and time it takes to train models etc .. at the moment it is limited to just the handful of wake words... but this will expand
did you flash the latest version or is still the one you tried from a month or so ago ?
also if you watch the ESPHome logs to see if you get any errors etc when it stops listening
Same issue described by these folks
After a few uses of the wakeword, you have to shout at it, and it does this in a loop
It scrolls off-page every 3 seconds or so
If I reset the device, it works normally and the log stops scrolling until you either wait a bit or use it a few times
using open wakeword the device is constantly streaming to HA in order to detect the wakeword. this will result in the logs constantly looping
Reset it. Working again for the moment
Sounds like the software is still very unstable
did you install the latest firmware ? or is this the older one that you flashed ?
Latest
try switching to on device and saying ok nabu as the wakeword and see if that gives the same result
I can try it for a while with on-device wakeword to see if that’s the unstable part
read my mind 😉
no problem , let me know how it goes. I'm in the UK so may be delay in responding at times
I have my Atom Echo working with wakewords (though it does behave a little oddly sometimes). If I can help getting PaulB's working in any way, please let me know!
i folowed this and its working however i want to use the wake work "ok jarvis" that i have set in the integration. i realize that i could switch it in the integration, however when i do it becomes unstable. it recognizes the new wake word but closed the connection before i could say anything
try changing the 'finished speaking detection' to 'relaxed' in the device page in the ESPHome integration in HA>>Settings>>devices and services>>esphome
Thanks I'll try it when I get home
The other thing I noticed, if it makes sense, when I switched to homeassistant before when I tried it, the white "idle" light turned a tealish color. Still turned blue to listen and then turned red to close the connection then back to that teal color
looking at the config, the colour of the idle 'listening' light changes depending on if the device is set to on device or home assistant wake word. My test device isnt in a case so it's hard to tell what the colours actually are so i am only going from the config 🙂
I've only seen my Atom Echo show red when HA is still loading the model, so I think it happens when there is a pipeline error.
Whew, that was a lot to go through, but didn't want to add a new post for an existing issue. I believe I have something quite similar. I have the pipline command stuff tested, but the atom devices don't work. I did discover that my PC would not work with voice assistant because it was logged into the local IP, not the secure external. Is there a way to check if the atoms are doing the same? One device worked one time and then stopped responding. Also, when I try to debug, there are options that say "Run Audio Pipeline" and "Run Audio Pipeline with Wake Word Detection" grayed out but I can "Run Text Pipeline". I can use audio assist from the PC or my phone.
in order to troubleshoot this the first place to start would be with ESPHome logs to make sure the device firmware is doing what it should be doing. If you can view the logs from the ESPHome dashboard and then paste them to a code share site and paste the link in here then we can take a look.
Please use a code share site to share code or logs, for example:
- https://dpaste.org/ (select YAML for the language, and consider picking a longer expiry)
- http://pastie.org/ (select YAML for the language)
- https://paste.debian.net/ (you guessed it, select YAML as the language)
Please don't use Pastebin, since it can randomly add spaces to the main view. Please also don't share text as images since it makes it harder for people to help you. Remember that others may have colour blindness, impaired vision, etc.
Same message again and again on one device, It looks like the IP may be the key. I think it should be trying my external address but don't know how to change the connection address or validate the connection (usually a web popup) WARNING m5stack-atom-echo-23864c @ 192.168.1.60: Connection error occurred: m5stack-atom-echo-23864c @ 192.168.1.60: The connection dropped immediately after encrypted hello; Try enabling encryption on the device or turning off encryption on the client (ESPHome Logs 2024.9.1).
The other atom is completely different https://dpaste.org/twOrg
@real talon Bump