1 paper · 1 filter
Lishi Zuo, Youzhi Tu, Lu Yi +4
Generative modeling provides a flexible way to model mixture-conditioned source distributions, but iterative diffusion and flow matching models are costly for long music signals. T…