Modules:
| Name | Description |
|---|---|
bigvgan_cuda_graph | CUDA graph wrapper for BigVGAN vocoding in IndexTTS2 Stage 1. |
codec | Inference-only EnhancedCodec runtime for IndexTTS 2.5. |
configuration_indextts2 | |
dit_cuda_graph | Per-shape CUDA graph capture for the S2Mel DiT transformer core. |
gpt | |
indextts2_s2mel_decoder | IndexTTS2 Stage 1: S2Mel decoder + BigVGAN vocoder. |
indextts2_talker | IndexTTS2 Stage 0: GPT-2 AR Talker with vLLM-native PagedAttention. |
pipeline | IndexTTS2 pipeline: GPT AR talker (text → mel codes) → S2Mel + BigVGAN (mel → audio). |
preprocess_utils | External model loading, audio I/O, and emotion conditioning for IndexTTS2. |
prompt_utils | Prompt helpers for IndexTTS2 talker prefill. |
s2mel | |
text_processing_v2_5 | Text preparation shared by IndexTTS 2.5 prompt sizing and model prefill. |
tokenizer | |
tokenizer_v2_5 | IndexTTS 2.5 multilingual tiktoken vocabulary. |
utils | |