We describe an end-to-end speech synthesis system that uses generative adversarial training. We train our Vocoder for raw phoneme-to-audio conversion, using explicit phonetic, pitch and duration modeling. We experiment with several pre-trained models for contextualized and decontextualized word embeddings and we introduce a new method for highly expressive character voice matching, based on discreet style tokens.
@article{arxiv.2310.09636,
title = {Generative Adversarial Training for Text-to-Speech Synthesis Based on Raw Phonetic Input and Explicit Prosody Modelling},
author = {Tiberiu Boros and Stefan Daniel Dumitrescu and Ionut Mironica and Radu Chivereanu},
journal= {arXiv preprint arXiv:2310.09636},
year = {2023}
}