IndexTTS 2.5: control the speed and pronunciation of a cloned voice

IndexTTS 2.5 from Bilibili clones a voice from one reference clip and speaks English, Chinese, Japanese, Spanish and Arabic. It has two controls I like for voiceovers: a speed factor from 0.5 to 2.0 (above 1 is slower, below 1 is faster), and pronunciation control, where you give it the reading of a word in English phonemes, Chinese Pinyin or Japanese Kana. You need an NVIDIA GPU with roughly 6 GB of memory. It is a custom license, and the reference voice needs permission. This guide shows you how to change the speed setting, test a difficult name, and check permission for the voice.