I believe Alexa devices work differently because the audio has to be sent to Amazon before being sent back down to the Echo (don’t get me started on Echo devices and Amazon’s data hoovering; suffice it to say don’t send anything you don’t want Amazon to aggressively use against you).
Config for atom echo with tts sent to separate media player.
Uses tts.cloud_say so will require home assistant cloud
Automatically sets the volume to 20% for voice responses and returns to previous volume state after response.
voice_assistant:
id: va
microphone: echo_microphone
speaker: echo_speaker
noise_suppression_level: 2
auto_gain: 31dBFS
volume_multiplier: 2.0
vad_threshold: 3
on_listening:
- light.turn_on:
id: led
blue: 100%
red: 0%
green: 0%
brightness: 100%
effect: pulse
on_tts_start:
- light.turn_on:
id: led
blue: 0%
red: 0%
green: 100%
brightness: 100%
effect: pulse
- homeassistant.service:
service: media_player.volume_set
data:
entity_id: media_player.living_room_speaker # Replace with your media player entity_id
volume_level: 0.2 # Set volume to 20%
- homeassistant.service:
service: tts.cloud_say
data:
entity_id: media_player.living_room_speaker
data_template:
message: "{{ tts_message }}"
variables:
tts_message: return x;
on_end:
- light.turn_off: led
- homeassistant.service:
service: media_player.volume_set
data_template:
entity_id: media_player.living_room_speaker
volume_level: "{{ states('media_player.living_room_speaker.attributes.volume_level') }}"
- delay: 100ms
- wait_until:
not:
speaker.is_playing:
- script.execute: reset_led
on_error:
- light.turn_on:
id: led
blue: 0%
red: 100%
green: 0%
Here’s a config that uses piper instead of tts.cloud_say for local tts
I run piper and faster-whisper via docker on a separate sever. You can simply use the piper addon if your hardware is decent or run as a container for very fast responses. As of 2024.12 my response times are a second or less.
voice_assistant:
id: va
microphone: echo_microphone
speaker: echo_speaker
noise_suppression_level: 2
auto_gain: 31dBFS
volume_multiplier: 2.0
vad_threshold: 3
on_listening:
- light.turn_on:
id: led
blue: 100%
red: 0%
green: 0%
brightness: 100%
effect: pulse
on_tts_start:
- light.turn_on:
id: led
blue: 0%
red: 0%
green: 100%
brightness: 100%
effect: pulse
- homeassistant.service:
service: media_player.volume_set
data:
entity_id: media_player.living_room_speaker # Replace with your media player entity_id
volume_level: 0.2 # Lower volume to 20%
- homeassistant.service:
service: tts.piper_say
data:
entity_id: media_player.living_room_speaker # Replace with your media player entity_id
message: "{{ tts_message }}"
variables:
tts_message: return x;
on_end:
- light.turn_off: led
- homeassistant.service:
service: media_player.volume_set
data_template:
entity_id: media_player.living_room_speaker
volume_level: "{{ states('media_player.living_room_speaker.attributes.volume_level') }}"
- delay: 100ms
- wait_until:
not:
speaker.is_playing:
- script.execute: reset_led
on_error:
- light.turn_on:
id: led
blue: 0%
red: 100%
green: 0%
For me, using this config the wake word works, but then it just gets stuck in “Assist satellite → Processing”, no errors in the logs. Before adding this to the config, the ATOM Echo was working fine, it only had the bad speaker sound.
I got: “homeassistant.exceptions.ServiceNotFound: Action tts.piper_say not found”
And: ‘M5Stack Atom Echo 23a170’ - No such effect ‘pulse’
I tried using:
- homeassistant.action:
service: tts.speak
data:
entity_id: tts.piper
media_player_entity_id: media_player.vlc_telnet
message: "{{ tts_message }}"
variables:
tts_message: return x;
The speaker works, but it just reads out loud “tts_message” literary, not the answer from the LLM. Any idea hot to fix it?
Edit:
Seems like using it like this works perfectly for me:
- homeassistant.action:
service: tts.speak
data:
entity_id: tts.piper
media_player_entity_id: media_player.vlc_telnet
message: !lambda 'return x;'
1 Like
thanks for posting. Where does this go though? in configuration.yaml or somewhere else? And the id, microphone, speaker names - where are these defined?
Thanks a lot in advance!
In the ESP device config file (ESPhome)
Thanks. I did the following -
- Installed ESPhome on docher (my HA instance also runs in docker, so no “addons” available)
- Configured the Atom Echo with HA firmware
- tried tinkering with the config file on ESPhome as shown above
substitutions:
name: m5stack-atom-echo-0ab278
friendly_name: M5Stack Atom Echo 0ab278
esphome:
name: ${name}
name_add_mac_suffix: false
friendly_name: ${friendly_name}
api:
encryption:
key:xxx
voice_assistant:
# Adjust Mic parameters for better understanding depending on room environment
noise_suppression_level: 2 # 1-4
volume_multiplier: 5.0 #1.0-6.0 (higher than 6 will result in major distortion)
# Don't use Atom's speaker at all
speaker: !remove
# Output response as a TTS to a chosen speaker
on_tts_start:
- homeassistant.service:
service: lms_tts_notify #this is what I otherwise use, works file via HA
data:
entity_id: media_player.black_2
message: !lambda 'return x;'
# I want to know when an Atom loses connection to HA, so blink the light fast red
on_client_disconnected:
then:
- voice_assistant.stop: {}
- micro_wake_word.stop: {}
- light.turn_on:
id: led
red: 1.0
green: 0.0
blue: 0.0
brightness: 1.0
effect: Fast Pulse
state: true
# Expose a restart button to HA so the Atom can be remotely rebooted
# (can fix a stuck pipeline or unstable wifi connection in multi-AP/mesh environments)
button:
- platform: restart
id: restart_btn
name: Reboot
disabled_by_default: false
icon: mdi:restart-alert
entity_category: config
device_class: restart
# Expose a new switch to HA to indicate timer_ringing AND ability to
# toggle it back to an off state (acknowledges the timer, same as pressing
# Atom's front button); automate using this switch
switch:
- platform: template
name: Timer Ringing
optimistic: true
lambda: |-
if (id(timer_ringing).state) {
return true;
} else {
return false;
}
turn_off_action:
- switch.turn_off: timer_ringing
wifi:
ssid: !secret wifi_ssid
password: !secret wifi_password
The problem - I see the following logs in the esphome
NFO ESPHome 2024.12.4
INFO Reading configuration /config/m5stack-atom-echo-0ab278.yaml...
Failed config
esphome: [source /config/m5stack-atom-echo-0ab278.yaml:5]
Platform missing. You must include one of the available platform keys: bk72xx, esp32, esp8266, host, libretiny, rp2040, rtl87xx.
name: m5stack-atom-echo-0ab278
name_add_mac_suffix: False
friendly_name: M5Stack Atom Echo 0ab278
So, am I on the correct path? Or what does this error actually mean?
Means you need something like this in your code:
esp32:
board: m5stack-atom
framework:
type: arduino
Thanks for the inputs. I got it working… almost.
only the “!remove”
is not accepted by the validate / compile function. I also have an ATOM. Do you know what the issue might be?
Thank you all. I got it working. Here is a complete “tutorial” for someone starting it from scratch like I did
Aim - to switch off the internal (“bad”) speaker of M5Stack Atom and route it to a media player (in my case an old Logitech Media Player radio)
My Configuration
- Homeassistant Docker 2025.1.4
- ESPhome installed also via docker (see configuration below)
- M5Stack Atom Echo
Homeassistant should be up and running. DO NOT follow the default guide for configuring M5Stack Atom on the HA website. This uses the internal speaker. It is not possible to change this at the current time via homeassistant. Therefore you need to have your own Esphome (server) instance
== Install ESPhome via Docker ==
ESPHome can be easily installed through docker
version: '3'
services:
esphome:
container_name: esphome
image: esphome/esphome
volumes:
- /volume1/docker/esphome/config:/config
- /etc/localtime:/etc/localtime:ro
restart: always
privileged: true
network_mode: host
If everything went well, you can access the dashboard at port 6052 of your host
http://192.168.1.XX:6052/
Connect your M5stackAtom to a computer and access the above link via Chrome to add this device to your dashboard.
==Firmware for M5StackAtom==
Here is a “correctly formatted” file which we can later compile and generate the required configuration.
The internal speaker has been commented out (line 65)
and homeassistant.service has been added via tts.google_say (to keep things simple and demonstrate). Change if you are using another tts service.
The mediaplayer to be used for the voice output is defined in entity_id
substitutions:
name: m5stack-atom-echo
friendly_name: M5Stack Atom Echo
micro_wake_word_model: okay_nabu # alexa, hey_jarvis, hey_mycroft are also supported
esphome:
name: ${name}
name_add_mac_suffix: true
friendly_name: ${friendly_name}
min_version: 2024.9.0
esp32:
board: m5stack-atom
framework:
type: esp-idf
version: 4.4.8
platform_version: 5.4.0
logger:
api:
encryption:
key: xxxxxxxxxxxx
ota:
- platform: esphome
id: ota_esphome
wifi:
ssid: !secret wifi_ssid
password: !secret wifi_password
captive_portal:
button:
- platform: factory_reset
id: factory_reset_btn
name: Factory reset
i2s_audio:
- id: i2s_audio_bus
i2s_lrclk_pin: GPIO33
i2s_bclk_pin: GPIO19
microphone:
- platform: i2s_audio
id: echo_microphone
i2s_din_pin: GPIO23
adc_type: external
pdm: true
speaker:
- platform: i2s_audio
id: echo_speaker
i2s_dout_pin: GPIO22
dac_type: external
channel: mono
voice_assistant:
id: va
microphone: echo_microphone
# speaker: echo_speaker
noise_suppression_level: 2
auto_gain: 31dBFS
volume_multiplier: 2.0
vad_threshold: 3
on_listening:
- light.turn_on:
id: led
blue: 100%
red: 0%
green: 0%
effect: "Slow Pulse"
on_stt_vad_end:
- light.turn_on:
id: led
blue: 100%
red: 0%
green: 0%
effect: "Fast Pulse"
on_tts_start:
- light.turn_on:
id: led
blue: 100%
red: 0%
green: 0%
brightness: 100%
effect: none
- homeassistant.service:
service: tts.google_say
data:
entity_id: media_player.black # Replace with your media player entity_id
#volume_level: 0.2 # Set volume to 20%
- homeassistant.service:
service: tts.google_say
data:
entity_id: media_player.black
data_template:
message: "{{ tts_message }}"
variables:
tts_message: return x;
on_end:
- delay: 100ms
- voice_assistant.stop:
- wait_until:
not:
voice_assistant.is_running:
- wait_until:
not:
switch.is_on: timer_ringing
- if:
condition:
lambda: return id(wake_word_engine_location).state == "On device";
then:
- micro_wake_word.start:
- script.execute: reset_led
else:
- voice_assistant.start_continuous:
- script.execute: reset_led
on_error:
- light.turn_on:
id: led
red: 100%
green: 0%
blue: 0%
brightness: 100%
effect: none
- delay: 2s
- script.execute: reset_led
on_client_connected:
- delay: 2s # Give the api server time to settle
- if:
condition:
lambda: return id(wake_word_engine_location).state == "On device";
then:
- micro_wake_word.start:
else:
- voice_assistant.start_continuous:
- script.execute: reset_led
on_client_disconnected:
- voice_assistant.stop:
- micro_wake_word.stop:
on_timer_finished:
- voice_assistant.stop:
- micro_wake_word.stop:
- switch.turn_on: timer_ringing
- wait_until:
not:
microphone.is_capturing:
- light.turn_on:
id: led
red: 0%
green: 100%
blue: 0%
brightness: 100%
effect: "Fast Pulse"
- while:
condition:
switch.is_on: timer_ringing
then:
- lambda: id(echo_speaker).play(id(timer_finished_wave_file), sizeof(id(timer_finished_wave_file)));
- delay: 1s
- wait_until:
not:
speaker.is_playing:
- light.turn_off: led
- switch.turn_off: timer_ringing
- if:
condition:
lambda: return id(wake_word_engine_location).state == "On device";
then:
- micro_wake_word.start:
- script.execute: reset_led
else:
- voice_assistant.start_continuous:
- script.execute: reset_led
binary_sensor:
# button does the following:
# short click - stop a timer
# if no timer then restart either microwakeword or voice assistant continuous
- platform: gpio
pin:
number: GPIO39
inverted: true
name: Button
disabled_by_default: true
entity_category: diagnostic
id: echo_button
on_multi_click:
- timing:
- ON for at least 50ms
- OFF for at least 50ms
then:
- if:
condition:
switch.is_on: timer_ringing
then:
- switch.turn_off: timer_ringing
else:
- if:
condition:
lambda: return id(wake_word_engine_location).state == "On device";
then:
- voice_assistant.stop
- micro_wake_word.stop:
- delay: 1s
- script.execute: reset_led
- script.wait: reset_led
- micro_wake_word.start:
else:
- if:
condition: voice_assistant.is_running
then:
- voice_assistant.stop:
- script.execute: reset_led
- voice_assistant.start_continuous:
- timing:
- ON for at least 10s
then:
- button.press: factory_reset_btn
light:
- platform: esp32_rmt_led_strip
id: led
name: None
disabled_by_default: true
entity_category: config
pin: GPIO27
default_transition_length: 0s
chipset: SK6812
num_leds: 1
rgb_order: grb
rmt_channel: 0
effects:
- pulse:
name: "Slow Pulse"
transition_length: 250ms
update_interval: 250ms
min_brightness: 50%
max_brightness: 100%
- pulse:
name: "Fast Pulse"
transition_length: 100ms
update_interval: 100ms
min_brightness: 50%
max_brightness: 100%
script:
- id: reset_led
then:
- if:
condition:
- lambda: return id(wake_word_engine_location).state == "On device";
- switch.is_on: use_listen_light
then:
- light.turn_on:
id: led
red: 100%
green: 89%
blue: 71%
brightness: 60%
effect: none
else:
- if:
condition:
- lambda: return id(wake_word_engine_location).state != "On device";
- switch.is_on: use_listen_light
then:
- light.turn_on:
id: led
red: 0%
green: 100%
blue: 100%
brightness: 60%
effect: none
else:
- light.turn_off: led
switch:
- platform: template
name: Use listen light
id: use_listen_light
optimistic: true
restore_mode: RESTORE_DEFAULT_ON
entity_category: config
on_turn_on:
- script.execute: reset_led
on_turn_off:
- script.execute: reset_led
- platform: template
id: timer_ringing
optimistic: true
internal: true
restore_mode: ALWAYS_OFF
on_turn_on:
- delay: 15min
- switch.turn_off: timer_ringing
select:
- platform: template
entity_category: config
name: Wake word engine location
id: wake_word_engine_location
optimistic: true
restore_value: true
options:
- In Home Assistant
- On device
initial_option: On device
on_value:
- if:
condition:
lambda: return x == "In Home Assistant";
then:
- micro_wake_word.stop
- delay: 500ms
- lambda: id(va).set_use_wake_word(true);
- voice_assistant.start_continuous:
- if:
condition:
lambda: return x == "On device";
then:
- lambda: id(va).set_use_wake_word(false);
- voice_assistant.stop
- delay: 500ms
- micro_wake_word.start
external_components:
- source: github://pr#5230
components:
- esp_adf
refresh: 0s
- source: github://jesserockz/esphome-components
components: [file]
refresh: 0s
esp_adf:
file:
- id: timer_finished_wave_file
file: https://github.com/esphome/wake-word-voice-assistants/raw/main/sounds/timer_finished.wav
micro_wake_word:
on_wake_word_detected:
- voice_assistant.start:
wake_word: !lambda return wake_word;
vad:
models:
- model: ${micro_wake_word_model}
Copy this file to the config of your esphome (in this case /volume1/docker/esphome/config).
Finally you should be seeing something like this -
== Validate, compile and download ==
Click on the “three dots” and check if the configuration is correct (“validate”). It works on my setup as pasted.
Now click “threedots” > Install. Select Manual Download. This will start the compilation and can take anywhere from 30 s to 5 mins depending on your host.
After successful compilation select the download file format
(I choose the older bin format) and download the file to your computer.
==Flash the M5Stack Atom Echo ==
start the download tool via this website (chrome)
https://web.esphome.io/?dashboard_install
and connect to your device. Select the correct serial port. You should be seeing a install link if the connection went through
Select the previously generated bin file and download. Wait for the process to end.
==Home Assistant ==
Add esphome integration in homeassistant. This will autodetect the esphome installed above. After about 2-3 minutes the device M5StackAtom will show up and work.
There are a few updates needed.
First to get rid of the “No such effect ‘pulse’” warning use effect: Slow Pulse instead.
And for the error regarding “voice_assistant > speaker”. Is because they moved from ‘speaker’ to ‘media_player’. So just use media_player: !remove
Hi all! Sorry I’ve been away a while. Yes, there have been a handful of changes, like speaker: being moved into media_player:…and now the new continuing conversations support, which is requiring some rework because TTS audio will often still be in the middle of playing on an HA media player and the ATOM will start listening again. This has resulted in some rather interesting “AI talking to itself” situations since I deployed 2025.5.0 today. (My local llama3 has a sarcastic personality, so it literally starts yelling and getting major attitude…with itself! 
)
I’ve also noticed that additional changes have resulted in media_player: !remove no longer removing the code for the speaker, even though it still doesn’t play (or maybe I just can’t hear that tiny speaker from my desk)?
Anyway, I’m getting a handle on the most recent changes and will rework my original examples soon. Since this is a cat/mouse game as ESPHome voice support evolves, I’m just going to host a copy of my latest yaml via git so these outdated examples stop tripping people up.
Well, media_player: !remove is actually screwing with the new conditionals for continuing conversation, as it is looking for the media_player to finish playing to continue listening. So, instead of using a !remove, I’m currently using
on_tts_start:
- media_player.volume_set: 0%
…followed by the - homeassistant.action: (homeassistant.service is being depreciated)
I haven’t found a way to base the conditional on the homeassistant.action completing because there’s no direct feedback via the API as to when the service call has completed…at least none that I’ve found yet without additional coding on HA’s side via a script or the like. So, technically, both my Atom and the tts.cloud_say run at the same time and the Atom uses the completion of the local playback to begin listening again, if necessary.
So, one problem squashed-ish. Not entirely ideal, as I’m still hearing a momentary crackle from the Atom speaker as the i2s pipeline kicks in, but it will have to do until another way to handling the start of the next listening round can be sorted out.
Yeah so when you comment out
media_player: echo_media_player
the follow-up question or continuing conversation doesn’t work anymore… Thats unfortunate, because i thought of a way to send the status of the media player to the atom with a helper. If we could make the pipeline wait for this event that would be awesome. However I tried everything but can’t get it to work…
Hi everyone, Thanks for your help it works great, my issue is if I want to play music in the speaker and used it to answer as voice assistant, the music stops playing and doesn’t restart. Do you have the same issue ? Did you find a solution for that ?
For anyone working with the ESP32-S3 AMOLED boards, I’ve put together a complete ESPHome YAML for the ESP32-S3 Touch AMOLED 1.75", including display, audio, and a local Home Assistant voice assistant setup.
GitHub repo here:
https://github.com/abramosphere/Home-Assistant-Voice-for-ESP32-S3-Touch-AMOLED-1.75
It’s hardware-specific but fully working, and might be useful as a reference (or starting point) for similar Waveshare / AMOLED S3 boards.
Does anyone have an up-to-date configuration using an original Atom Echo with either local or docker-based wakeword and an external HA media player with the speaker disabled on the echo itself?
Have been tinkering a fair amount but my echo currently stops working if I disable the speaker and otherwise just stalls for ages (presumably buffer issues) and then won’t recognise another wake word even after it eventually partially plays the response that my HA media player played a minute or two prior!