VoxBox AUTO

An automatic melody-to-model music box generator

VoxBox AUTO is a C Group Project Extension in First Year Computing at Imperial College London. The project ran from May to June 2026. The project takes a user's sung or played melody and automatically generates a 3D-printable music box mechanism to play that exact tune.

As an independent extension (not using any pre-provided code from the department), this repository contains the pipeline: frontend, conversion, and modelling.

Award: Selected as a group of top 12 extensions to present at Imperial College London Open Days and to Arm Holdings.

---

Contributors

Ordered in alphabetical order by last name

• Bert Chun Chang (lead)
• Isaac Chi
• Wania Siddiqui
• Derek Wen

---

Project Structure

├── Makefile            // Build rules for compilation  
├── conversion.c        // Sample conversion routines used by pipeline  
├── conversion.h        // Header for conversion.c  
├── dywapitchtrack.c    // Pitch tracking implementation  
├── dywapitchtrack.h    // Header for dywapitchtrack.c  
├── frontend.py         // Logic for Python frontend  
├── metronome.c         // Logic for timing/metronome  
├── metronome.h         // Header for metronome.c  
├── savefile.c          // Functions to write out .wav file  
├── savefile.h          // Header for savefile.c  
├── stl_generator.c     // Generates STL file with input CSV file  
├── stl_generator.h     // Header for stl_generator.c  
├── tempo.c             // Logic for changing the recording tempo  
├── tempo.h             // Header for tempo.c  
├── tinywav.c           // Utilises tinywav library to read .wav files  
├── tinywav.h           // Header for tinywav.c  
└── voxbox_auto.py      // Describes the UI for the Python frontend  

---

Features & Pipeline

Our pipeline is broken down into several modular components spanning audio capture, signal processing, and 3D generation:

1. PyQt5 application to record user input melodies.
2. Stores the captured microphone data as a .wav file.
3. Converts the .wav file into a .csv file (note, angle), utilising C libraries (see below).
4. Uses vector arithmetic to convert the .csv file into a .stl music box cylinder model.
5. Logic linking the .stl file with the 3D printing software.
6. Designing the hardware and music box mechanism.

---

Running the Project

Note that you need a 3D printer using BambuLab to print the model.

1. Clone the repository.
2. Run voxbox_auto.py.

---

Note on GitHub Usage

During VoxBox AUTO development, we used the Git services on the GitLab website provided by Imperial College London. As a result, the commit history is not detailed on GitHub.

---

Acknowledgments

This project uses the following open-source libraries for audio processing:

• dywapitchtrack (Antoine Schmitt)
  • Real-time dynamic wavelet algorithm (dywa) pitch tracking to determine note frequencies.
  • Distributed under the MIT License.
• TinyWav (Martin Roth)
  • Reading and writing .wav audio files directly from and to memory.
  • Distributed under the ISC License.