MusicGAN
lastmod 2026-11-10created 2020-01-093 edits
MusicGAN is a name for models that apply generative adversarial networks (GANs) to music generation. The generator produces score representations or audio waveforms and the discriminator learns to tell them apart from real music; as with WaveGAN, which generates waveforms directly, the key challenge is how to represent the temporal structure of music. The link on this page is an implementation presented at a TensorFlow hackathon.