1 paper · 1 filter
Ke Gu, Zhicong Wu, Peng Bai +5
Existing singing voice synthesis (SVS) models largely rely on fine-grained, phoneme-level durations, which limits their practical application. These methods overlook the complement…