IndexTTS2 is an open-source zero-shot text-to-speech (TTS) model capable of generating realistic human voices without the need for speaker-specific training data. It separates speaker identity from emotional tone, allowing you to fully control emotion, prosody, and timing for each utterance.