Overview
MERIC is a music generation and retrieval research project built around a shared music semantic anchor.
The core idea is to separate multimodal understanding from acoustic synthesis. Text, image, or video inputs are mapped into a music-aligned semantic space, then a generative audio model turns that anchor into sound.
Research Shape
MERIC connects two tasks that are often treated separately:
- generation — producing music from multimodal input,
- retrieval — finding music through the same semantic space.
That shared representation makes the system more than a one-way generator. It becomes a bridge between musical meaning and musical sound.
Public Status
This project is listed here as a public-facing research dossier. The source repository is private, so the site does not sync its README or expose repository internals.