Neural Speech Synthesis

Presented during the IRCAM Forum @NYU 2022

In this talk, Nicolas Obin and Axel Roebel from the Sound Analysis and Synthesis (AS) team will present their latest research on neural speech synthesis with a particular focus on three axis: speech synthesis using neural vocoder, neural voice identity conversion with few-shot learning, and neural speech emotion transformation.

The talk will be illustrated using numerous examples including the vocal deep fake reconstruction of past personalities, such as the French comedian and singer Dalida or the father of science-fiction Isaac Asimov.