🗣️ Request a New Wake Word

You can request a new microWakeWord model by opening a GitHub issue.

✅ How to request a word

1. Go to the Issues tab
2. Click New issue
3. Set the title to:

mww: your wake word here

Examples:

mww: hey tater  
mww: tater totterson  
mww: hello computer  

That’s it — no labels, no templates, no body text required.

---

🔄 What happens next

• The .tflite and .json files are added to the repository
• The issue is labeled, commented on, and closed when complete

---

⚠️ Notes

• Test your wake word with TTS first.
  Make sure your text-to-speech engine pronounces the phrase the way you expect.
  You may need to spell it phonetically or a little “funny” so TTS says it correctly — the trainer uses the same pronunciation.
• Please request one wake word per issue
• Avoid punctuation or emojis in the title
• Training runs sequentially if multiple requests are open

---

🗣️ Set Up Your Custom Wake Word on Home Assistant Voice

⚠️ Important: voicePE-TaterTimer.yaml is for Voice PE, but the same structure and steps apply to any Home Assistant voice device.
You can mimic these instructions for your own hardware by updating the equivalent file for your device.

All of the settings below are located at the very top of the YAML file inside the substitutions: section.
You no longer need to hunt for line numbers — everything commonly edited lives in one place.

Open voicePE-TaterTimer.yaml (or your device’s YAML) and edit the substitutions: block.

---

🧾 Device Name & Friendly Name

Change how the device appears in ESPHome and Home Assistant:

device_name: tatervpe
friendly_name: TaterVPE

---

📡 Wi-Fi & Network Settings

Set your Wi-Fi credentials (or use secrets) and optionally pin the device to a Home Assistant Voice IP:

wifi_ssid: !secret wifi_ssid
wifi_password: !secret wifi_password
ha_voice_ip: "127.0.0.1"

If you don’t want a fixed IP, simply remove ha_voice_ip and the device will use DHCP.

---

🎙️ Wake Word Model

Choose the wake word model and give it a matching ID:

wake_word_name: hey_tater
wake_word_model_url: https://raw.githubusercontent.com/TaterTotterson/microWakeWords/refs/heads/main/microWakeWords/hey_tater.json

The wake_word_name must match the model ID used internally.

---

🎚️ Wake Word Sensitivity

Tune how sensitive the wake word detection is:

wake_cutoff_slight: "250"     # Slightly sensitive (very strict)
wake_cutoff_moderate: "245"   # Balanced
wake_cutoff_very: "222"       # Very sensitive

Lower numbers = more sensitive
Higher numbers = fewer false activations

---

🔔 Optional – Change the Wake Sound

You can customize the sound played when the wake word is detected.

Edit the wake sound URL in the substitutions section:

wake_word_triggered_sound_file: https://github.com/esphome/home-assistant-voice-pe/raw/dev/sounds/wake_word_triggered.flac

You can point this to any compatible .mp3 or .flac file hosted online.

---

✅ Final Notes

• These values are read throughout the config automatically
• No other parts of the YAML need to be edited
• Test your wake word in TTS first to ensure it’s pronounced correctly
(you may need to spell it creatively for best results)