UC ITExpo Project

StudyBuddy is a portable lecture recorder built on a Raspberry Pi 5. It captures audio through an I2S MEMS microphone, transcribes speech in real time on-device, and sends recordings to a server that extracts structured notes, key terms, assignments, and schedule items using a local LLM.

The client is designed to run on a Raspberry Pi 5 with:

• 3.5" touchscreen (480×320) — the UI is fixed to this resolution and runs fullscreen with the cursor hidden
• IMS I2S MEMS microphone — connected via the Pi's I2S interface for audio capture

Add the following to /boot/firmware/config.txt:

dtparam=i2s=on
dtoverlay=googlevoicehat-soundcard

Then reboot. Verify the mic is detected:

arecord -l

You may also need to set it as the default capture device in /etc/asound.conf:

pcm.!default {
  type asym
  capture.pcm "mic"
}
pcm.mic {
  type plug
  slave {
    pcm "hw:1,0"
  }
}

flask
flask-cors
openai-whisper
requests

pip install flask flask-cors openai-whisper requests

Whisper also requires ffmpeg:

# Ubuntu/Debian (including Raspberry Pi OS)
sudo apt install ffmpeg

Install Ollama from https://ollama.com, then pull the model:

ollama pull llama3.2:3b

pyaudio
vosk
Pillow

pip install pyaudio vosk Pillow

PyAudio also requires PortAudio:

sudo apt install portaudio19-dev

The client uses Vosk for on-device live transcription. Download the small English model and place it next to recorder_app.py:

wget https://alphacephei.com/vosk/models/vosk-model-small-en-us-0.15.zip
unzip vosk-model-small-en-us-0.15.zip

The expected folder name is vosk-model-small-en-us-0.15.

Start Ollama in a separate terminal on the server machine:

ollama serve

Start the server:

cd Server
python3 server.py

Open http://localhost:5000 in a browser to access the web interface.

Start the client on the Raspberry Pi:

cd Client
python3 recorder_app.py

The app will launch fullscreen on the 3.5" display. Make sure the server IP is configured in recorder_app.py before running.

• Matthew Gentle
• Ryan Parker