Drumify

Inspiration

In an age where AI-assisted tooling is prevalent, we wanted to enable people to have direct agency in making music in an intuitive way.

What it does

It records your voice and turns it into a drum loop consisting of kicks, hi-hats, and snares.

How we built it

We wrote it in C++ as a Juce plugin.

Challenges we ran into

Wrestling with messy audio input and getting it into a format that was easy to classify was tough.

Accomplishments that we're proud of

Implementing proper onset detection, a decent heuristic-based classification algorithm, and integrating the whole pipeline.

What we learned

How to use Juce. We designed and implemented a real-time audio processing pipeline in C++ using JUCE, including envelope-based onset detection, feature extraction (RMS and zero-crossing rate), heuristic drum classification, and offline loop rendering. The system operates directly on live microphone input and produces a synchronized drum loop without relying on pretrained models or external libraries. Integrating all components into a stable, low-latency plugin was a major technical milestone.