Inspiration
We wanted to make music creation feel instant, playful, and accessible. Most tools assume you already have sounds or know production. We flipped that: what if anything around you, like a water bottle or desk tap, could become the music itself?
What it does
GPT wRAPPER is a browser based beat making app that turns everyday sounds into full tracks. You record short snippets, trigger them like a sample pad, and choose a style such as trap, lo-fi, drill, or house. The backend arranges your clips into a structured beat and can resynthesize melodies, replacing traditional instruments with your custom sound while preserving pitch and timing.
How we built it
We built a web app that captures audio directly in the browser and slices it into playable clips. On the backend, we process and arrange those clips into rhythmic patterns based on the selected style. For melody resynthesis, we map the pitch contour of an existing melody onto a single recorded sample, effectively turning it into a playable instrument.
Challenges we ran into
One of the hardest parts was making random sounds actually sound musical. Simple playback sounded messy, so we had to carefully handle timing, pitch alignment, and clipping. Resynthesizing melodies while preserving both rhythm and pitch without distortion was also tricky. Latency between recording and playback needed to be minimal to keep the experience fun.
Accomplishments that we're proud of
We turned completely unstructured noise into something that feels like real music. The ability to take a single sound and recreate a melody with it is especially satisfying. We are also proud that everything runs in the browser, making it fast and accessible with no setup.
What we learned
We learned a lot about NLP audio processing, especially pitch detection, time alignment, and sound manipulation. We also learned how important user experience is in creative tools. Even small delays or awkward controls can break the flow.
What's next for GPT wRAPPER
We want to expand the range of styles and improve the quality of the generated beats. Adding more control over arrangement, layering multiple sounds, and enabling sharing or collaboration are next steps. We are also exploring ways to make the resynthesis more expressive and dynamic.
Built With
- css
- fastapi
- javascript
- json
- midi-llm
- next.js
- python
- react
- typescript
Log in or sign up for Devpost to join the conversation.