AI Assistant (Piper) setup

Piper is M2C2’s voice assistant. Say “Hey Piper” (wake phrase may be configurable) followed by a command. Piper ties into calendar, reminders, smart home, sports, health, and system actions.

1. Microphone hardware

Plug a USB microphone into a USB 2/3 port (avoid flaky passive hubs).
Position the mic below or beside the mirror frame to reduce glass reflections.
Run the wizard mic test or open Settings → Voice → Input level.

💡 Tip: See the USB microphone guide for shopping notes and gain staging.

2. Wake word & sensitivity

Open Settings → Voice → Wake word.
Choose sensitivity Medium first.
If you get false activations (TV dialog, conversations), lower sensitivity or move the mic.
If Piper misses you, raise sensitivity slightly or reduce room echo (rug/curtains).

3. Speech-to-text (STT)

Depending on your build:

On-device — lower latency, works offline for some commands, higher CPU use.
Cloud — better accuracy for complex sentences; requires internet.

Pick the mode in Settings → Voice → Recognition.

ℹ️ Note: Cloud STT sends short audio clips to your configured provider. Read the privacy sheet for your release before enabling.

4. Text-to-speech (TTS)

Open Settings → Voice → Piper TTS.
Choose voice pack / language.
Adjust speech rate and volume so replies are clear over bathroom fans or open kitchens.

5. Command permissions

Restrict dangerous actions (for example unlock, purchase, delete) if your build exposes them — use Settings → Voice → Permissions.

⚠️ Warning: Voice assistants can be triggered accidentally. Do not wire voice commands to door locks or alarms without strong confirmation prompts.

Troubleshooting

No waveform in mic test
Wrong input device selected; try alsamixer / OS sound settings on the Pi.

Piper hears but does nothing
Check PM2 logs for NLP errors; confirm internet if using cloud STT.