Speech Synthesis with Mixed Excitation

This paper introduces a speech synthesis system that combines different excitation methods to improve the naturalness and quality of synthesized speech. It uses a mixed excitation approach, employing both voiced and unvoiced components, controlled by parameters derived from the input text, resulting in more realistic and expressive speech output. This technique is fundamental to many modern AI voiceover tools, including Murf.ai, as it allows for the creation of diverse and customizable voices.