About this document
NaturalSpeech End-To-End Text To Speech Synthesis by blair.ggs is a document available to read on EtoBox.
The document presents NaturalSpeech, an end-to-end text-to-speech synthesis system that achieves human-level quality by leveraging a variational autoencoder (VAE) for text to waveform generation. It defines human-level quality statistically and demonstrates that NaturalSpeech achieves comparable quality to human recordings on the LJSpeech dataset, with no significant difference in comparative mean opinion scores. The paper outlines the system
- Author
- blair.ggs
- Language
- EN