Index-TTS is an industrial-level, controllable zero-shot text-to-speech system that introduces precise duration control into an autoregressive TTS architecture while preserving natural prosody from voice prompts. It achieves disentanglement of speaker identity and emotional expression, enabling independent control over timbre and emotion through audio reference, emotional vector specification, or free-text description, with support for Chinese and English and optional FP16 and DeepSpeed acceleration.