Vidu: China's First Long-Duration, High-Consistency, High-Dynamic Video Large Model 366

Vidu is a video large model developed by Shengshu Technology in collaboration with Tsinghua University. With its innovative U-ViT architecture, it has become the first domestic video large model to reach the Sora level. This model not only features global technical innovation, but also demonstrates its large-scale scalability through the open-source UniDiffuser model.

Vidu can not only simulate the real physical world, but also possesses rich imagination, supporting multi-shot generation and high spatiotemporal consistency. The Vidu model integrates Diffusion and Transformer technologies, with the innovative development of the U-ViT architecture. It is capable of generating high-definition video content up to 16 seconds long and up to 1080P resolution with a single click.

Vidu Features:

• Model Architecture: The Vidu model integrates Diffusion and Transformer technologies, with the innovative development of the U-ViT architecture.

• Video Generation Capability: Vidu is capable of generating high-definition video content up to 16 seconds long and up to 1080P resolution with a single click.

• 3D Generation: Based on simple text input, generate high-quality, diverse, and creative images. Image Generation API calls, online experience. Based on text or image input, generate personalized 3D models | Based on existing 3D models, automatically edit textures and more.
  

• Practical Application Demonstration: The official materials showcase a video example featuring "a ship in a studio sailing toward the camera," demonstrating the realistic effects of the ship and waves.

• Advanced Features: Vidu can not only simulate the real physical world, but also possesses rich imagination, supporting multi-shot generation and high spatiotemporal consistency.

Beijing Shengshu Technology Co., Ltd. (referred to as "Shengshu Technology") was established in March 2023. The core team members come from the Institute for Artificial Intelligence Research at Tsinghua University. In addition, the company brings together top talents from well-known tech companies such as Alibaba, Tencent, and ByteDance, making it a leading deep generative algorithm research team worldwide, with the capability of underlying innovative R&D for diffusion probabilistic models. The company is committed to building a world-leading multimodal large model, integrating multimodal information such as text, images, video, and 3D, exploring the commercial empowerment of generative AI in scenarios such as art design, game production, post-production of film and television, and content social networking, and enhancing human creativity and productivity through AI.