Shengshu unveils vidu s2 with real-time video editing and VR compatibility

On September 15, 2026, Shengshu Technology launched the Vidu S2 video large model series, featuring two core models: Vidu S2-Avatar and Vidu S2-Editing. Vidu S2-Avatar enables continuous interaction with digital characters through text or voice, supporting dynamic additions of items, clothing, or background images. It also enhances real-time output resolution to 720P. Vidu S2-Editing allows real-time video editing based on text instructions and reference images, handling tasks like style transfer and virtual try-on. Shengshu’s Zhang Jintao led the development, emphasizing innovations in training methods and performance optimization. The Vidu S2 series integrates real-time avatar output into spatial video generation, supporting VR headset compatibility. It also introduced Self-Replay Forcing (SRF) training and lightweight Refiner technology to improve efficiency. In evaluations, S2-Avatar achieved top results in nine indicators, while S2-Editing scored 4.26 in joint assessments. Shengshu noted that spatial video remains in the fixed perspective stage, with future challenges including reducing latency and exploring panoramic formats.