Overview
SoundDrop is a campus music map for generating and publishing place-based music.
A user chooses a location on the map, uploads an image, writes a short memory or prompt, and receives a small set of generated music versions. The chosen version can be published to the public map or saved privately.
The project treats music generation as a local, social object. A song is attached to a corner, a path, a building, a photo, and a moment.
Product Shape
SoundDrop is built around a compact flow:
- choose a place on a campus map,
- add visual and textual context through an image and short story,
- generate music through the MERIC backend,
- select, download, and save one version,
- publish publicly or keep it private.
The map is the interface. Existing works stay visible as context while new pieces are being placed, so the site feels like a growing layer of campus memory.
System
The public product combines a React frontend, an Express API, PostgreSQL storage, object uploads, and a GPU inference service running MERIC.
The open-source surface is the frontend UI. The production backend and model infrastructure remain private, while the public UI repository documents the interaction layer and visual system.
Direction
SoundDrop is an experiment in making generated music feel situated.
Instead of treating output as isolated files, it gives music a place, a prompt, an image, and a publishing decision. The result is closer to a living sound map than a generation demo.