The folder contains the VAE encoder and the VAE decoder used in the paper Using a manifold vocoder for spectral voice and style conversion . The encoder and decoder was trained on the multi-speaker TIMIT database.

We also provide the samples for the samples of the three experiments in the paper: vocoding evaluation (samples/VOD), voice conversion evaluation ( samples/{F2F, F2M, M2F, M2M}), and style conversion evaluation (samples/SC).

F2F stands for female-to-female mapping; F2M stands for female-to-male mapping; M2F stands for male-to-female mapping; M2M stands for male-to-male mapping. The file name has a format of srcSpeaker-trgSpeaker_sentID_mappingFeat.wav.