BizyAir IndexTTS 2.5 Text to Speech is an industrial-level zero-shot text-to-speech system that clones a voice from a single reference audio clip. It supports 5 languages (Chinese, English, Japanese, Spanish, Arabic) with cross-lingual and timbre-emotion disentanglement, fine-grained emotion control via emo_vector / emo_audio_prompt / emo_text, speaking speed control via duration_factor (0.5x–2.0x), and pronunciation control (Pinyin / CMU phonemes / Japanese Kana). Features faster inference than IndexTTS 2, BF16 inference, and vLLM production deployment. 0.8B params, returns wav/mp3/flac.
