GANymede

Trying some implementations of GANs for audio in Keras.

WaveGAN

WaveGAN implements a structure that enables working on raw audio.

• Paper : Synthesizing Audio with Generative Adversarial Networks

Advancement

• Generator structure
• Discriminator structure
• Loss function : WGAN-GP - (Experimenting some issues)
• Train function
• Datasets :
• • Speech Commands
• • Speech Commands Zero through Nine (SC09). Source : here
• • Drum sound effects. Source : here
• • Bach piano performances. Source : here

GANSynth

• Paper : GANSynth: Adversarial Neural Audio Synthesis

Incoming implementation