Deep Voice 3: Scaling Text-to-Speech with Convolutional Sequence Learning

We present Deep Voice 3, a fully-convolutional attention-based neural-to-speech (TTS) system. Deep Voice 3 matches state-of-the-art neural synthesis systems in naturalness while training ten times faster. We Deep Voice 3 to data set sizes unprecedented for TTS, training on more eight hundred hours of audio from over two thousand speakers. In addition, identify common error modes of attention-based speech synthesis networks, how to mitigate them, and compare several different waveform methods. We also describe how to scale inference to ten million per day on one single-GPU server.