Today we release “PodcastMix”, a dataset for separating music and speech in podcasts:

• website
• arXiv paper
• trained separation models at GitHub
• download dataset at Zenodo