IndexTts2DurationConditioning
Enums · AiDotNet.TextToSpeech.Latest
How IndexTTS2’s text-to-semantic model receives the target length.
Fields
| Field | Summary |
|---|---|
IndexTts2DurationConditioning Paper |
The paper’s (§ Duration Control): one token p = W_sem h(T) after the conditioning, the target length’s row of the semantic position table (tied), or zeros for free-length generation. |
IndexTts2DurationConditioning ReleaseSpeedTokens |
The released checkpoint’s: two learned tokens (speed_emb rows 1 then 0) after the conditioning and no length control. |
int value__ |