vschaos is an open-source Python library for variational neural audio synthesis. This library, based on pytorch, allows high-level functions for creating, training, and using many variants of variational auto-encoding. It provides :

  • Easy data loading, with a unified dataset architecture, with methods specifically targeting audio data handling (spectral transforms, raw waveform slicing, transformations)
  • Modular definition of variational auto-encoders and generative adversarial networks, allowing to define custom hybridations
  • Easy definition for model conditioning and semi-supervised methods
  • Algebraic losses, allowing to define custom training criteria
  • Max patches for real-time / offline generation
  • Toy dataset generation methods for direct experimentation (see toy datasets here)
  • Provides high

This toolbox is open-source, and can be found on github.

Course Syllabus