StyleTts2TrainingStage

Enums · AiDotNet.TextToSpeech.StyleEmotion

The modules a StyleTTS 2 training step updates (Li et al. 2023, §3.3 and Appendix C).

Fields

Field Summary
StyleTts2TrainingStage Joint Stage 2 (train_second.py): PL-BERT, the prosody predictor and prosodic style encoder; then the style diffusion and the wave discriminators; then joint fine-tuning of the decoder and acoustic style encoder with SLM adversarial training and…
StyleTts2TrainingStage Pretraining Stage 1 (the reference’s train_first.py): waveform reconstruction from the aligned phonemes, the extracted pitch and energy and the acoustic style — the text encoder, acoustic style encoder and decoder, and once the aligner’s objectives s…
int value__