Deep generative models for raw audio synthesis
Audio synthesis is a challenging task in speech processing with a history of more than 50 years. Following the progress in other areas, speech community started to adapt deep learning approaches for raw audio generation. Now, despite the success of such models as WaveNet, efficient high-quality audio synthesis remains an active area of research. Dmytro will present an overview of the current state of things in the area and share some practical insights on desing and optimization of deep generative models with applications in audio processing.