Talking Fish

Wisconsin Robotics Outreach 2025/2026

Folder Structure

• Voice: AI voice assistant
• Camera: CV facial tracking
• MotorControl: Servo controls

Package Manager

This project's dependencies are managed by uv. The pyproject.toml file contains the dependencies list, and descriptions of basic commands can be found here.

• uv run program.py: Runs your program with up-to-date dependencies.
• uv add: Add a dependency to the entire project. (This updates pyproject.toml and runs uv sync.)
• uv remove: Remove a dependency from the entire project. (This updates pyproject.toml and runs uv sync.)
• uv sync: Updates your environment to match pyproject.toml.
• Install uv for development on personal computer

AI Voice Response

1. Connect the Bluetooth speaker (with microphone) to the Pi.

2. Run: uv run Voice/main.py

   PiperTTS source code

File Organization

Voice/
├── main.py ├── voice/
│ ├── config.py (For constants, settings, and Finley's prompt)
│ ├── audio_io.py (Microphone & speaker I/O)
│ ├── health.py (Startup device checks)
│ ├── stt_whisper.py (Speech to text)
│ ├── llm_ollama.py (LLM response generation)
│ ├── tts_piper.py (Text to speech with streaming)
│ └── app.py (Main loop)
├── en_GB-northern_english_male-medium.onnx (TTS voice model)
└── en_GB-northern_english_male-medium.onnx.json (TTS voice model)

To Do

• Add activation trigger (Speak "Hey Finley" or press a button)
• Decrease Time To First Token (Start speech to text before LLM generation is completed, switch to a faster (smaller) LLM, test the AI HAT we ordered)
• Give Finley a personality!!!! 🐟🐟🐟
  • Allow mechanical responses? (e.g., Finley says "--shake_head--" and robot shakes its head.)

Camera & CV Face Tracking

Servo Controls