AI-Generated Music Detection
Overview
The field studies how to distinguish AI-generated music from human-made recordings, including tracks where only some stems or production stages involve AI. Detectors look for acoustic artifacts and codec residuals, analyze learned audio embeddings or token spaces, model the distribution of real music, and combine features tied to musical content. Evaluation increasingly tests transfer to unseen generators and codecs, as well as robustness to editing, broadcast conditions, and mixed human-AI production. A central challenge is that cues that work on clean, familiar examples may fail under real-world shifts, while binary labels cannot express how much AI contributed.
Latest papers
16 papers1 paper in the last four weeks, down 67% on the four weeks before. 0.0% of all new papers.