A voice-activated nighttime note recorder for Raspberry Pi. Record your thoughts at night without grabbing your phone. Transcriptions are delivered to your phone the next morning.

• Wake word activation - Say "Porcupine" or "Jarvis" to start recording
• Automatic transcription - Uses OpenAI Whisper API
• Phone delivery - Sends notes via Telegram or Email
• Hands-free operation - No screens, no phones, just voice
• Local storage - All recordings saved on the Pi

• Raspberry Pi 4 Model B
• Microphone (USB mic recommended, or headphones with inline mic)
• microSD card (8GB+)
• Power supply

• Raspberry Pi OS (Bullseye or later)
• Python 3.8+
• Internet connection

cd ~
mkdir voice-nightlight
cd voice-nightlight
# Copy all project files here

chmod +x install.sh
./install.sh

Edit the .env file with your credentials:

nano .env

Required API Keys:

1. Go to https://console.picovoice.ai/
2. Sign up for free account
3. Create an access key
4. Copy key to PICOVOICE_ACCESS_KEY in .env

1. Go to https://platform.openai.com/api-keys
2. Create new API key
3. Copy key to OPENAI_API_KEY in .env
4. Cost: ~$0.006 per minute of audio

1. Open Telegram and message @BotFather
2. Send: /newbot
3. Follow instructions, copy your bot token
4. Add to TELEGRAM_BOT_TOKEN in .env
5. Message @userinfobot to get your chat ID
6. Add to TELEGRAM_CHAT_ID in .env
7. Start a chat with your new bot

For Gmail:

1. Enable 2FA on your Google account
2. Create App Password: https://support.google.com/accounts/answer/185833
3. Use app password in SMTP_PASSWORD

# List audio devices
arecord -l

# Record 5 second test
arecord -d 5 -f cd test.wav

# Play it back
aplay test.wav

If your microphone isn't working:

# Check audio configuration
alsamixer

# Select the right input device
sudo nano /etc/asound.conf

source venv/bin/activate
python3 voice_nightlight.py

You should see:

🌙 VOICE NIGHTLIGHT - Ready for bedtime notes
👂 Listening for wake word: 'porcupine'...

1. Say: "Porcupine" (wait for beep/confirmation)
2. Speak your note
3. Say: "Stop recording" or press Ctrl+C
4. Wait for transcription and delivery
5. Check your Telegram or email!

To make it run automatically when Pi boots:

chmod +x setup_autostart.sh
sudo ./setup_autostart.sh

Control the service:

sudo systemctl start voice-nightlight   # Start
sudo systemctl stop voice-nightlight    # Stop
sudo systemctl status voice-nightlight  # Check status
sudo journalctl -u voice-nightlight -f  # View logs

• ✅ Instant notifications
• ✅ Easy setup
• ✅ Works everywhere
• ✅ Free

• ✅ Universal
• ⚠️ May be delayed
• ⚠️ Might go to spam

Wake Words:

• "Porcupine" (easier)
• "Jarvis" (sounds cooler)

Recording:

• Speak naturally, 1-2 feet from mic
• USB mic works better than headphone mic
• Quiet room = better transcription

Stopping:

• Say "stop recording"
• Or just press Ctrl+C on the Pi
• Auto-stops after 5 minutes

voice-nightlight/
├── voice_nightlight.py    # Main application
├── requirements.txt       # Python dependencies
├── .env                   # Your API keys (DO NOT COMMIT)
├── .env.example          # Template
├── install.sh            # Setup script
├── setup_autostart.sh    # Autostart configuration
├── recordings/           # Audio files (created automatically)
│   ├── recording_20240108_235959.wav
│   └── recording_20240108_235959.txt
└── README.md            # This file

# Check if mic is detected
arecord -l

# If using USB mic, try:
sudo nano /boot/config.txt
# Add: dtparam=audio=on

# Reboot
sudo reboot

• Check your API key in .env
• Make sure there are no extra spaces
• Get new key at https://console.picovoice.ai/

• Verify your API key
• Check you have credits: https://platform.openai.com/usage
• First $5 is free for new accounts

• Make sure you started a chat with your bot
• Verify chat ID with @userinfobot
• Check bot token is correct

• Use a better microphone (USB recommended)
• Reduce background noise
• Speak clearly, not too fast
• Try changing language in code (language="en")

• Raspberry Pi 4: ~€50 (one-time)
• USB Microphone: ~€15 (one-time)
• OpenAI Whisper: ~€0.006/minute
• Picovoice: Free tier (1 wake word)
• Telegram: Free

Example: 5 minutes of recording per night = €0.03/night = €1/month

For mass production, consider:

1. Hardware:

   • ESP32-S3 with built-in WiFi (~€5)
   • I2S MEMS microphone (~€2)
   • LED nightlight PCB
   • 3D printed enclosure
   • USB-C power

2. Software:

   • Edge ML for wake word (no API cost)
   • Cloud backend for transcription
   • Mobile app for setup & delivery
   • OTA updates

3. Features:

   • Multiple user profiles
   • Smart scheduling (only active at night)
   • Voice activity detection
   • Better stop word detection
   • Data encryption

MIT License - Feel free to use for personal or commercial projects!

This is an MVP. Ideas for improvement:

• Better stop phrase detection (Vosk)
• Local Whisper (faster-whisper)
• iOS Shortcuts integration
• Multi-language support
• Voice activity detection
• Cloud sync
• Mobile app
• LED status indicators

• Place the device 1-2 meters from your bed
• Use a soft nightlight (not bright)
• Speak at normal volume
• Don't stress about perfect transcription
• Review notes in the morning, not immediately

Made with 💙 for better sleep and captured ideas

Questions? Issues? Open an issue or contribute!