Soul AI Lab Open Podcast Synthesis Model SoulX-Podcast

The Soul App AI team's official open-source voice synthesis model SoulX-Podcast, which supports multi-lingual dialects such as the Sino-Enchuan Sing, can stabilize the output of over 60-minute multi-wheel voice dialogues; the model supports a multi-wheel dialogue capability for zero sample cloning, which can achieve cross-square acoustic cloning, with natural voice with a dialect character based on Qwen3-1.7B as the base, using the LLM + Flow Matching voice generation paradigm, and achieves the best results in the viewer environment with a degree of similarity to the sound。