GptSoVitsStage

Enums · AiDotNet.TextToSpeech.CodecBased

Which of GPT-SoVITS’s networks GPTSoVITS trains.

Fields

Field Summary
GptSoVitsStage FlowMatching s2 for v3 to V5: the flow-matching synthesizer, by default LoRA on the DiT’s attention (s2_train_v3_lora.py; LoraRank 0 trains it all, as s2_train_v3.py).
GptSoVitsStage SoVits s2 for v1 to v2ProPlus: the SoVITS synthesizer and its discriminators as a GAN (s2_train.py).
GptSoVitsStage TextToSemantic s1: the text-to-semantic GPT (s1_train.py; ScaledAdam).
int value__