MOSS-TTSD

Expressive multi-speaker spoken-dialogue TTS (v0/v0.5), Qwen3-based, with zero-shot two-speaker voice cloning and generation up to 960 seconds — released "by Fudan University @Open_MOSS" from Xipeng Qiu's group. The later MOSS-TTS family (2026, 800K+ HF downloads) migrated to OpenMOSS (SII)/MOSI.AI and is tracked under SII.