Paper

PRVAE-VC: Non-parallel many-to-many voice conversion with perturbation-resistant variational autoencoder

Kou Tanaka, Hirokazu Kameoka, Takuhiro Kaneko

Audio samples for ATRCORPORA

(Supported: Safari, Chrome, FireFox, Opera)

Source speaker: mho (male) mht (male) mmy (male) msh (male) mtk (male) myi (male) fkn (female) fks (female) ftk (female) fym (female) , Target speaker: mho (male) mht (male) mmy (male) msh (male) mtk (male) myi (male) fkn (female) fks (female) ftk (female) fym (female) , Utterance: 01 02 03

Re-synthesized	Source
	Target
Converted	CVAE-Small (1240k params)
	+ beta
	+ cc (cycle consistency)
	+ ac (auxiliary classifier)
	+ pr (perturbation resistance, ours)
Converted	CVAE-Large (8842k params)
	+ beta
	+ cc (cycle consistency)
	+ ac (auxiliary classifier)
	+ pr (perturbation resistance, ours)