Audio Synthesis by Generative Adversarial Network (GAN)

Has anyone worked in audio synthesis using GAN?

I am trying to implement the matlab code using magnitude and phase information (e.g. 128x128x2), which means I have two images one for magnitude and one for phase. Mainly, GANSynth work. The spectrogram based implementation MATLAB already provides. I want to check the results for my audio dataset (not music) using both Magnitude and Phase spectrograms. Any help would be highly appreciated.

Answer 1

[jibrahim]

Hi Hina,

Please check this exaample:

Train Generative Adversarial Network (GAN) for Sound Synthesis

https://www.mathworks.com/help/audio/ug/train-gan-for-sound-synthesis.html;jsessionid=13f059d539541a16eddd3721a7b7

It only uses magnitude information (and reconstructs the phase), but it should get you started.