The First Domestic Self-Developed Video Large Model Vidu Launched

TapTechNews August 1st news, the first domestic purely self-developed video large model Vidu has been launched, and currently has opened two core functions of text-to-video and image-to-video to users, providing two duration options of 4 seconds and 8 seconds, and the highest resolution is 1080P. Users can experience it without application, just register with an email.

TapTechNews attaches the Vidu official website: www.vidu.studio

The Vidu model first debuted at the 2024 Zhongguancun Forum in April this year, jointly released by Tsinghua University and Shengshu Technology. Its core technology, the U-ViT architecture, was proposed by the team in September 2022, earlier than the DiT architecture used by Sora, and it is the world's first architecture that combines Diffusion and Transformer.

After more than 3 months of improvement, Vidu has added capabilities such as character consistency (CharacterToVideo), anime style, generation of text and special effect pictures.

According to the introduction, Vidu can achieve near-cinema-level effects in aspects such as composition, narration, and light and shadow. Vidu can also generate film and television-level special effect pictures, such as smoke, glare effects, CG special effects, etc.