Depuis quelques jours, j’essaie de configurer un ESP32 et un ESP32-S3 en tant qu’assistant vocal. Pour cela, j’utilise un microphone INMP441.
Voici la conf de mes deux esp:
esphome:
name: assistant-chambre
friendly_name: Assistant Chambre
platformio_options:
board_build.flash_mode: dio
on_boot:
- light.turn_on:
id: led_ww
blue: 100%
brightness: 60%
effect: fast pulse
esp32:
board: esp32-s3-devkitc-1
variant: esp32s3
framework:
type: esp-idf
sdkconfig_options:
CONFIG_ESP32S3_DEFAULT_CPU_FREQ_240: "y"
CONFIG_ESP32S3_DATA_CACHE_64KB: "y"
CONFIG_ESP32S3_DATA_CACHE_LINE_64B: "y"
CONFIG_AUDIO_BOARD_CUSTOM: "y"
psram:
mode: octal
speed: 80MHz
# Enable logging
logger:
level: DEBUG
ota:
wifi:
ssid: !secret wifi_ssid
password: !secret wifi_password
manual_ip:
# Set this to the IP of the ESP
static_ip: 192.168.1.23
# Set this to the IP address of the router. Often ends with .1
gateway: 192.168.1.1
# The subnet of the network. 255.255.255.0 works for most home networks.
subnet: 255.255.255.0
# Enable fallback hotspot (captive portal) in case wifi connection fails
ap:
# Enable Home Assistant API
api:
on_client_connected:
then:
- delay: 50ms
- light.turn_off: led_ww
- micro_wake_word.start:
on_client_disconnected:
then:
- voice_assistant.stop:
button:
- platform: restart
name: "Restart"
id: but_rest
switch:
- platform: template
id: mute
name: mute
optimistic: true
on_turn_on:
- micro_wake_word.stop:
- voice_assistant.stop:
- light.turn_on:
id: led_ww
red: 100%
green: 0%
blue: 0%
brightness: 60%
effect: fast pulse
- delay: 2s
- light.turn_off:
id: led_ww
- light.turn_on:
id: led_ww
red: 100%
green: 0%
blue: 0%
brightness: 30%
on_turn_off:
- micro_wake_word.start:
- light.turn_on:
id: led_ww
red: 0%
green: 100%
blue: 0%
brightness: 60%
effect: fast pulse
- delay: 2s
- light.turn_off:
id: led_ww
light:
- platform: esp32_rmt_led_strip
id: led_ww
rgb_order: GRB
pin: GPIO48
num_leds: 1
rmt_channel: 0
chipset: ws2812
name: "on board light"
effects:
- pulse:
- pulse:
name: "Fast Pulse"
transition_length: 0.5s
update_interval: 0.5s
min_brightness: 0%
max_brightness: 100%
# Audio and Voice Assistant Config
i2s_audio:
- id: i2s_mic
i2s_lrclk_pin: GPIO3
i2s_bclk_pin: GPIO2
- id: i2s_spk
i2s_lrclk_pin: GPIO6
i2s_bclk_pin: GPIO7
microphone:
platform: i2s_audio
id: va_mic
adc_type: external
i2s_audio_id: i2s_mic
i2s_din_pin: GPIO4
channel: left
pdm: false
speaker:
platform: i2s_audio
id: va_spk
dac_type: external
i2s_audio_id: i2s_spk
i2s_dout_pin: GPIO8
micro_wake_word:
on_wake_word_detected:
- voice_assistant.start:
- light.turn_on:
id: led_ww
red: 30%
green: 30%
blue: 70%
brightness: 60%
effect: fast pulse
model: hey_jarvis
voice_assistant:
id: va
microphone: va_mic
speaker: va_spk
noise_suppression_level: 2.0
volume_multiplier: 4.0
on_stt_end:
then:
- light.turn_off: led_ww
on_error:
- micro_wake_word.start:
on_end:
then:
- light.turn_off: led_ww
- wait_until:
not:
voice_assistant.is_running:
- micro_wake_word.start:
esphome:
name: test
friendly_name: test
esp32:
board: esp32dev
framework:
type: arduino
version: recommended
# Enable logging
logger:
level: DEBUG
# Enable Home Assistant API
api:
ota:
wifi:
ssid: !secret wifi_ssid
password: !secret wifi_password
manual_ip:
# Set this to the IP of the ESP
static_ip: 192.168.1.19
# Set this to the IP address of the router. Often ends with .1
gateway: 192.168.1.1
# The subnet of the network. 255.255.255.0 works for most home networks.
subnet: 255.255.255.0
# Enable fallback hotspot (captive portal) in case wifi connection fails
ap:
captive_portal:
i2s_audio:
- id: i2s_in
i2s_lrclk_pin: GPIO26 #WS
i2s_bclk_pin: GPIO25 #SCK
microphone:
- platform: i2s_audio
adc_type: external
pdm: false
id: mic_i2s
channel: left
bits_per_sample: 32bit
i2s_audio_id: i2s_in
i2s_din_pin: GPIO33 #SD Pin from the INMP441 Microphone
voice_assistant:
microphone: mic_i2s
id: va
noise_suppression_level: 2
auto_gain: 31dBFS
volume_multiplier: 4.0
use_wake_word: false
on_wake_word_detected:
- light.turn_on:
id: led_light
on_listening:
- light.turn_on:
id: led_light
effect: "Scan Effect With Custom Values"
red: 63%
green: 13%
blue: 93%
on_stt_end:
- light.turn_on:
id: led_light
effect: "None"
red: 0%
green: 100%
blue: 0%
on_error:
- light.turn_on:
id: led_light
effect: "None"
- if:
condition:
switch.is_on: use_wake_word
then:
- switch.turn_off: use_wake_word
- delay: 1sec
- switch.turn_on: use_wake_word
on_tts_start: # this is required to play the output on a media player
- homeassistant.service:
service: tts.speak
data:
media_player_entity_id: media_player.marantz_sr6015 #replace this with your media player entity id
message: !lambda 'return x;'
entity_id: tts.piper #replace this with your piper tts id.
on_client_connected:
- if:
condition:
switch.is_on: use_wake_word
then:
- voice_assistant.start_continuous:
on_client_disconnected:
- if:
condition:
switch.is_on: use_wake_word
then:
- voice_assistant.stop:
on_end:
- light.turn_off:
id: led_light
binary_sensor:
- platform: status
name: API Connection
id: api_connection
filters:
- delayed_on: 1s
on_press:
- if:
condition:
switch.is_on: use_wake_word
then:
- voice_assistant.start_continuous:
on_release:
- if:
condition:
switch.is_on: use_wake_word
then:
- voice_assistant.stop:
switch:
- platform: template
name: Use wake word
id: use_wake_word
optimistic: true
restore_mode: RESTORE_DEFAULT_ON
entity_category: config
on_turn_on:
- lambda: id(va).set_use_wake_word(true);
- if:
condition:
not:
- voice_assistant.is_running
then:
- voice_assistant.start_continuous
on_turn_off:
- voice_assistant.stop
- lambda: id(va).set_use_wake_word(false);
light:
- platform: neopixelbus
id: led_light
type: grb
pin: GPIO32 # DIN pin of the LED Strip
num_leds: 9 # change the Number of LEDS according to your LED Strip.
name: "Light"
variant: SK6812
default_transition_length: 0.5s
effects:
- addressable_scan:
name: Scan Effect With Custom Values
move_interval: 50ms
scan_width: 2
Home Assistant, OpenWakeWord, Whisper et Piper sont installés sur Docker. Lorsque je teste l’ensemble du processus avec mon téléphone en mode debug, le mot de réveil et les commandes prononcés sont détectés. Sur mes ESP, le mot de réveil n’est pas détecté.
J’ai ensuite testé microWakeWord sur mon ESP32-S3 et cette fois-ci cela a marché.
Par contre, c’est la partie Whisper qui ne semble pas fonctionner plus. On dirait presque un soucis de connexion entre mes esp et ha/openwakeword/whisper/piper.
Je ne comprends pas ce qui se passe et je suis à court d’idées. Quelqu’un pourrait-il m’indiquer d’où vient le problème, s’il vous plaît ?
Bonjour @Nicolas117
J’utilise les mêmes solutions que @WarC0zes, le S3 très réactif mais je n’en ai sorti aucun retour son, si ça fonctionne pour toi je suis preneur de ta conf stp
Bob
Merci @WarC0zes, mais avant de complexifier les choses j’aimerais faire fonctionner mes esp.
Ma pipeline fonctionne plutôt bien lorsque j’utilise mon téléphone (whisper répond en 2 à 3s).
Le problème semble vraiment être du coté d’esphome.
J’ai eu beau faire tous les tutos du web, rien ne fonctionne. Je désespère .
Je suis avec un i3 sans gpu pour le conteneur whisper avec le model small-int8.
Ok je vais tester vosk mais je n’ai pas vu de doc sur le hub docker pour l’installation de l’image. De ce que je comprends sur le github il faut que je télécharge le modèle français sur le conteneur et que je passe le model utilisé en argument ?
Pour les logs de mes esp:
[20:23:14][I][app:102]: ESPHome version 2024.2.2 compiled on Mar 19 2024, 20:22:36
[20:23:14][C][wifi:577]: WiFi:
[20:23:14][C][wifi:409]: Local MAC: A0:B7:65:63:24:7C
[20:23:14][C][wifi:414]: SSID: [redacted]
[20:23:14][C][wifi:415]: IP Address: 192.168.1.19
[20:23:14][C][wifi:417]: BSSID: [redacted]
[20:23:14][C][wifi:418]: Hostname: 'test'
[20:23:14][C][wifi:420]: Signal strength: -50 dB ▂▄▆█
[20:23:14][C][wifi:424]: Channel: 11
[20:23:14][C][wifi:425]: Subnet: 255.255.255.0
[20:23:14][C][wifi:426]: Gateway: 192.168.1.1
[20:23:14][C][wifi:427]: DNS1: 0.0.0.0
[20:23:14][C][wifi:428]: DNS2: 0.0.0.0
[20:23:14][C][logger:447]: Logger:
[20:23:14][C][logger:448]: Level: DEBUG
[20:23:14][C][logger:449]: Log Baud Rate: 115200
[20:23:14][C][logger:451]: Hardware UART: UART0
[20:23:14][C][template.switch:068]: Template Switch 'Use wake word'
[20:23:14][C][template.switch:091]: Restore Mode: restore defaults to ON
[20:23:14][C][template.switch:057]: Optimistic: YES
[20:23:14][C][captive_portal:088]: Captive Portal:
[20:23:14][C][mdns:115]: mDNS:
[20:23:14][C][mdns:116]: Hostname: test
[20:23:14][C][ota:096]: Over-The-Air Updates:
[20:23:14][C][ota:097]: Address: 192.168.1.19:3232
[20:23:14][C][ota:100]: Using Password.
[20:23:14][C][ota:103]: OTA version: 2.
[20:23:14][C][api:139]: API Server:
[20:23:14][C][api:140]: Address: 192.168.1.19:6053
[20:23:14][C][api:142]: Using noise encryption: YES
[20:25:13][D][switch:012]: 'Use wake word' Turning ON.
[20:25:13][D][switch:055]: 'Use wake word': Sending state ON
[20:25:13][D][voice_assistant:416]: State changed from IDLE to START_PIPELINE
[20:25:13][D][voice_assistant:422]: Desired state set to START_MICROPHONE
[20:25:13][D][voice_assistant:118]: microphone not running
[20:25:13][D][voice_assistant:202]: Requesting start...
[20:25:13][D][voice_assistant:416]: State changed from START_PIPELINE to STARTING_PIPELINE
[20:25:13][D][voice_assistant:437]: Client started, streaming microphone
[20:25:13][D][voice_assistant:416]: State changed from STARTING_PIPELINE to START_MICROPHONE
[20:25:13][D][voice_assistant:422]: Desired state set to STREAMING_MICROPHONE
[20:25:13][D][voice_assistant:155]: Starting Microphone
[20:25:13][D][voice_assistant:416]: State changed from START_MICROPHONE to STARTING_MICROPHONE
[20:25:13][D][voice_assistant:523]: Event Type: 1
[20:25:13][D][voice_assistant:526]: Assist Pipeline running
[20:25:13][D][voice_assistant:416]: State changed from STARTING_MICROPHONE to STREAMING_MICROPHONE
[20:25:13][D][voice_assistant:523]: Event Type: 9
[20:25:15][D][switch:016]: 'Use wake word' Turning OFF.
[20:25:15][D][switch:055]: 'Use wake word': Sending state OFF
[20:25:15][D][voice_assistant:516]: Signaling stop...
[20:25:15][D][voice_assistant:416]: State changed from STREAMING_MICROPHONE to STOP_MICROPHONE
[20:25:15][D][voice_assistant:422]: Desired state set to IDLE
[20:25:15][D][voice_assistant:416]: State changed from STOP_MICROPHONE to STOPPING_MICROPHONE
[20:25:15][D][voice_assistant:523]: Event Type: 0
[20:25:15][E][voice_assistant:653]: Error: no_wake_word - No wake word detected
[20:25:15][D][voice_assistant:516]: Signaling stop...
[20:25:15][D][voice_assistant:416]: State changed from STOPPING_MICROPHONE to STOP_MICROPHONE
[20:25:15][D][voice_assistant:422]: Desired state set to IDLE
[20:25:15][D][voice_assistant:416]: State changed from STOP_MICROPHONE to IDLE
[20:25:15][D][voice_assistant:523]: Event Type: 2
[20:25:15][D][voice_assistant:613]: Assist Pipeline ended
[20:25:21][D][switch:012]: 'Use wake word' Turning ON.
[20:25:21][D][switch:055]: 'Use wake word': Sending state ON
[20:25:21][D][voice_assistant:416]: State changed from IDLE to START_PIPELINE
[20:25:21][D][voice_assistant:422]: Desired state set to START_MICROPHONE
[20:25:21][D][voice_assistant:118]: microphone not running
[20:25:21][D][voice_assistant:202]: Requesting start...
[20:25:21][D][voice_assistant:416]: State changed from START_PIPELINE to STARTING_PIPELINE
[20:25:21][D][voice_assistant:437]: Client started, streaming microphone
[20:25:21][D][voice_assistant:416]: State changed from STARTING_PIPELINE to START_MICROPHONE
[20:25:21][D][voice_assistant:422]: Desired state set to STREAMING_MICROPHONE
[20:25:21][D][voice_assistant:155]: Starting Microphone
[20:25:21][D][voice_assistant:416]: State changed from START_MICROPHONE to STARTING_MICROPHONE
[20:25:21][D][voice_assistant:523]: Event Type: 1
[20:25:21][D][voice_assistant:526]: Assist Pipeline running
[20:25:21][D][voice_assistant:416]: State changed from STARTING_MICROPHONE to STREAMING_MICROPHONE
[20:25:21][D][voice_assistant:523]: Event Type: 9
[20:25:57][D][esp32.preferences:114]: Saving 1 preferences to flash...
[20:25:57][D][esp32.preferences:143]: Saving 1 preferences to flash: 0 cached, 1 written, 0 failed
[20:27:58][I][ota:117]: Boot seems successful, resetting boot loop counter.
[20:27:58][D][esp32.preferences:114]: Saving 1 preferences to flash...
[20:27:58][D][esp32.preferences:143]: Saving 1 preferences to flash: 0 cached, 1 written, 0 failed
Et pour le s3 avec microwakeword :
[22:22:32][C][wifi:577]: WiFi:
[22:22:32][C][wifi:409]: Local MAC: 48:27:E2:D3:0D:A0
[22:22:32][C][wifi:414]: SSID: [redacted]
[22:22:32][C][wifi:415]: IP Address: 192.168.1.23
[22:22:32][C][wifi:417]: BSSID: [redacted]
[22:22:32][C][wifi:418]: Hostname: 'assistant-chambre'
[22:22:32][C][wifi:420]: Signal strength: -39 dB ▂▄▆█
[22:22:32][C][wifi:424]: Channel: 11
[22:22:32][C][wifi:425]: Subnet: 255.255.255.0
[22:22:32][C][wifi:426]: Gateway: 192.168.1.1
[22:22:32][C][wifi:427]: DNS1: 0.0.0.0
[22:22:32][C][wifi:428]: DNS2: 0.0.0.0
[22:22:32][D][light:036]: 'on board light' Setting:
[22:22:32][D][light:085]: Transition length: 1.0s
[22:22:32][W][micro_wake_word:150]: Wake word is already running
[22:22:32][C][logger:447]: Logger:
[22:22:32][C][logger:448]: Level: DEBUG
[22:22:32][C][logger:449]: Log Baud Rate: 115200
[22:22:32][C][logger:451]: Hardware UART: USB_SERIAL_JTAG
[22:22:32][C][esp32_rmt_led_strip:175]: ESP32 RMT LED Strip:
[22:22:32][C][esp32_rmt_led_strip:176]: Pin: 11
[22:22:32][C][esp32_rmt_led_strip:177]: Channel: 0
[22:22:32][C][esp32_rmt_led_strip:202]: RGB Order: GRB
[22:22:32][C][esp32_rmt_led_strip:203]: Max refresh rate: 0
[22:22:32][C][esp32_rmt_led_strip:204]: Number of LEDs: 1
J’ai ajouté porcupine et vosk. Vosk est plus rapide ± 0.3s avec le modèle small mais est un peu moins précis que whisper.
En ce qui concerne mes esp, les résultats sont malheureusement les mêmes. Le wake word n’est pas détecté et coté stt il ne se passe rien.
INFO ESPHome 2024.2.2
INFO Reading configuration /config/test.yaml...
INFO Starting log output from 192.168.1.19 using esphome API
INFO Successfully connected to test @ 192.168.1.19 in 0.211s
INFO Successful handshake with test @ 192.168.1.19 in 0.099s
[08:02:47][I][app:102]: ESPHome version 2024.2.2 compiled on Mar 20 2024, 07:57:22
[08:02:47][C][wifi:577]: WiFi:
[08:02:47][C][wifi:409]: Local MAC: A0:B7:65:63:24:7C
[08:02:47][C][wifi:414]: SSID: [redacted]
[08:02:47][C][wifi:415]: IP Address: 192.168.1.19
[08:02:47][C][wifi:417]: BSSID: [redacted]
[08:02:47][C][wifi:418]: Hostname: 'test'
[08:02:47][C][wifi:420]: Signal strength: -57 dB ▂▄▆█
[08:02:47][C][wifi:424]: Channel: 11
[08:02:47][C][wifi:425]: Subnet: 255.255.255.0
[08:02:47][C][wifi:426]: Gateway: 192.168.1.1
[08:02:47][C][wifi:427]: DNS1: 0.0.0.0
[08:02:47][C][wifi:428]: DNS2: 0.0.0.0
[08:02:47][C][logger:447]: Logger:
[08:02:47][C][logger:448]: Level: DEBUG
[08:02:47][C][logger:449]: Log Baud Rate: 115200
[08:02:47][C][logger:451]: Hardware UART: UART0
[08:02:47][C][light:103]: Light 'Light'
[08:02:47][C][light:105]: Default Transition Length: 0.5s
[08:02:47][C][light:106]: Gamma Correct: 2.80
[08:02:47][C][template.switch:068]: Template Switch 'Use wake word'
[08:02:47][C][template.switch:091]: Restore Mode: restore defaults to ON
[08:02:47][C][template.switch:057]: Optimistic: YES
[08:02:47][C][status:034]: Status Binary Sensor 'API Connection'
[08:02:47][C][status:034]: Device Class: 'connectivity'
[08:02:47][C][captive_portal:088]: Captive Portal:
[08:02:47][C][mdns:115]: mDNS:
[08:02:47][C][mdns:116]: Hostname: test
[08:02:47][C][ota:096]: Over-The-Air Updates:
[08:02:47][C][ota:097]: Address: 192.168.1.19:3232
[08:02:47][C][ota:100]: Using Password.
[08:02:47][C][ota:103]: OTA version: 2.
[08:02:47][C][api:139]: API Server:
[08:02:47][C][api:140]: Address: 192.168.1.19:6053
[08:02:47][C][api:142]: Using noise encryption: YES
[08:02:58][D][switch:012]: 'Use wake word' Turning ON.
[08:02:58][D][switch:055]: 'Use wake word': Sending state ON
[08:02:58][D][voice_assistant:416]: State changed from IDLE to START_PIPELINE
[08:02:58][D][voice_assistant:422]: Desired state set to START_MICROPHONE
[08:02:58][D][voice_assistant:118]: microphone not running
[08:02:58][D][voice_assistant:202]: Requesting start...
[08:02:58][D][voice_assistant:416]: State changed from START_PIPELINE to STARTING_PIPELINE
[08:02:58][D][voice_assistant:437]: Client started, streaming microphone
[08:02:58][D][voice_assistant:416]: State changed from STARTING_PIPELINE to START_MICROPHONE
[08:02:58][D][voice_assistant:422]: Desired state set to STREAMING_MICROPHONE
[08:02:58][D][voice_assistant:155]: Starting Microphone
[08:02:58][D][voice_assistant:416]: State changed from START_MICROPHONE to STARTING_MICROPHONE
[08:02:58][D][voice_assistant:523]: Event Type: 1
[08:02:58][D][voice_assistant:526]: Assist Pipeline running
[08:02:58][D][voice_assistant:416]: State changed from STARTING_MICROPHONE to STREAMING_MICROPHONE
[08:02:58][D][voice_assistant:523]: Event Type: 9
[08:03:18][D][esp32.preferences:114]: Saving 1 preferences to flash...
[08:03:18][D][esp32.preferences:143]: Saving 1 preferences to flash: 0 cached, 1 written, 0 failed
Quand j’active le debug sur ha au niveau de la pipeline assist, avec mon esp-32s3 qui a microwakeword d’activé le .wav destiné à whisper ou vosk fait 44 octets.
De même avec l’esp32 pour le wakword et porcupine/openwakeword lorsque que je coupe l’utilisation du wakeword via le switch.
Je ne comprends vraiment pas ce qui est incorrect dans ma configuration.
.
