Preprint
Aug 2026
Beyond Dry References: Learning Relative Audio Effects Representations via Contrastive Distance Learning
RelFx is proposed, a contrastive learning framework that learns relative effect transformations from general audio collections without requiring dry references during representation training, and demonstrates state-of-the-art performance under the standard Fx-Encoder++ MUSDB18 evaluation protocol.
Xinlu Liu, Huibin Lin, Weixing Wei et al.
· 0 citations