Reproduce “MixJudge: Learning a Language-Queryable Critic for Vocal Mixing” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Kuan-Hsun Ho, Yuan-Chiao Cheng, Yen-Tung Yeh, Yi-Hsuan Yang
MixJudge evaluates named problems in vocal mixes and compares processed alternatives. It represents vocal, backing, and blend alongside language descriptions, and trains on synthetic degradations to judge issue identity and severity. Across 88 menus, its mean-evidence policy selected a version acoustically closer to the held-out professional mix than no edit in 72.7% of cases; this measures distance, not listener preference.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.