OpenSep: Leveraging Large Language Models with Textual Inversion for Open World Audio Separation

Tanvir Mahmud, Diana Marculescu

Audio separation in real-world scenarios, where mixtures contain a variable number of sources, presents significant challenges due to limitations of existing models, such as over-s…