Audio samples of the results would be nice, but I didn't find any links in the pdf My guess would be that extracting useful audio is possible, but the idea evo

Audio samples of the results would be nice, but I didn't find any links in the pdf My guess would be that extracting useful audio is possible, but the idea evoked by the paper's title "Deep Karaoke", of flawlessly removing vocals from a finished song, is a lot harder, or maybe impossible. Anyway this has many useful applications