Zero-shot expressive voice cloning and speech generation. Generate anything from short clips to full-length audiobooks with realistic emotional delivery, pacing, and breath control. Clone any voice from a 10-second reference and perform emotions the original speaker never recorded.
MOSS‑TTS Family is an open‑source speech and sound generation model family from MOSI.AI and the OpenMOSS team. It is designed for high‑fidelity, high‑expressiveness, and complex real‑world scenarios, covering stable long‑form speech, multi‑speaker dialogue, voice/character design, environmental sound effects, and real‑time streaming TTS.
Qwen3-TTS is an open-source series of TTS models developed by the Qwen team at Alibaba Cloud, supporting stable, expressive, and streaming speech generation, free-form voice design, and vivid voice cloning.
