IndexTts2Stage

Enums · AiDotNet.TextToSpeech.Latest

Which of IndexTTS2’s networks IndexTTS2 trains.

Fields

Field Summary
IndexTts2Stage Emotion Text-to-semantic stage 2: [c + e, …] with the speaker conditioner frozen, the emotion conditioner trained and a speaker classifier behind a gradient reversal layer pushing speaker identity out of e (Eq.
IndexTts2Stage FineTune Text-to-semantic stage 3: every conditioner frozen, the rest fine-tuned.
IndexTts2Stage SemanticToMel The semantic-to-mel flow matching (Eq.
IndexTts2Stage TextToSemantic Text-to-semantic stage 1: the whole model on [c, p, BT, text, BA, sem], the duration token zeroed with DurationDropProbability.
int value__