Audio style transfer

This audio style transfer experiment uses neural synthesis applied to audio waveforms, coupled with face interpolation video generated by reversible generative model GLOW. The synthesized videos are played after each pair of original audio.