I have some Atom Echos installed as HA remote voice assistants. They’re very cool, but they seem to say “I’m sorry I didn’t understand that” a bit too often when I’m not addressing them.

The Echos are thinking I’m giving them commands when there’s a discussion between people in the room or when a show/movie/music is playing. I have a custom wakeword, but I don’t think any sounds happening in the background are sounding like it - there is only 1 word in English that rhymes with the first part of my wake word.

So I’m wondering if there’s a way to configure the Echos to be more strict on what they consider to be the wakeword, or to be less attentive to ambient sound (or to require a more direct command, like “WAKEWORD” said kind of loud).

  • Mike Wooskey@lemmy.d.thewooskeys.comOPEnglish
    1·
    8 months ago

    I enabled the assist_pipeline and retrieved and listened to the audio files from my Echos, but when I tried to look at the esphome/m5stack-atom-echo-wake-word.yaml file to edit the values for noise_suppression_level, auto_gain, or volume_multiplier, the file doesn’t exist. I do have an esphome directory and it contains m5stack-atom-echo-xxxxxx.yaml files for each of my echos, but inside those .yaml files there is no voice_assistant section.

    Can you please paste the contents of your m5stack-atom-echo-wake-word.yaml file (obfuscating anything private, of course)? I’ll try manually creating this file to see if the Echos recognize it.