Fish Speech (Voice Cloning)

Fish Speech is a multilingual text-to-speech and zero-shot voice cloning system from Fish Audio that reproduces a target voice from a short reference clip. The current Fish Audio S2 generation uses a dual-autoregressive architecture with reinforcement-learning alignment, trained on a large multilingual corpus for natural, emotionally varied speech across many languages. The open-source release is distributed under the Apache 2.0 license.