Deep Learning with Keras

WaveNet — a generative model for learning how to produce audio

WaveNet is a deep generative model for producing raw audio waveforms. This breakthrough technology was introduced (https://deepmind.com/blog/wavenet-generative-model-raw-audio/) by Google DeepMind (https://deepmind.com/) for teaching users how to speak to computers. The results are truly impressive, and you can find online examples of synthetic voices where the computer learns how to talk with the voices of celebrities such as Matt Damon. So, you might wonder why learning to synthesize audio is so difficult. Well, each digital sound we hear is based on 16,000 samples per second (sometimes, 48,000 or more), and building a predictive model where we learn to reproduce a sample based ...