Kuaishou Launches Future High-Tech Video Intelligence Innovation Center Showcasing AI, Codec, and Transmission Advances

On March 1, Kuaishou held the inaugural ceremony of the Future High-Tech Video Intelligence Innovation Center at its Beijing campus, bringing together officials from the Beijing Municipal Science and Technology Commission, Zhongguancun Administrative Committee, the Municipal Economic and Information Commission, Haidian District Government, as well as scholars from top universities and research institutes.

The center, selected in 2022 as one of Beijing's first enterprise technology innovation hubs, receives special support from the municipal science commission and partners with Tsinghua University, Renmin University, the National Radio and Television Administration’s Planning Institute, and the China Academy of Information and Communications Technology’s Cloud Computing and Big Data Institute to drive disruptive video‑intelligence research.

Key technical achievements highlighted include the self‑developed KVC video codec, which reduces bitrate by over 50% compared to the open‑source X265 while maintaining or improving visual quality, the SL200 encoding chip that cuts bitrate by nearly 20% at equal quality, and the new CMTP transmission protocol that significantly lowers playback stalling and is being standardized with the China Communications Standards Association.

Kuaishou also demonstrated its 6DoF free‑viewpoint video technology showcased at the Asian Games fencing events and the Kuaishou Village BA, as well as its digital human “Guan Xiaofang,” which has over 1.8 million followers and has been used for virtual employee applications.

In the AI domain, Kuaishou’s large models now span 13 billion, 66 billion, and 175 billion parameters, with the 175 billion version approaching GPT‑4 performance; its visual generation models rival MidJourney V5, and the company has deployed AI bots like “AI Xiao Kuai” for interactive comment‑section assistance.

The company’s R&D efforts have produced more than 100 papers at top international conferences, earning awards such as the best paper at CIKM 2022 and a best‑paper candidate at SIGIR 2023, while maintaining deep collaborations with over 20 leading universities worldwide and investing tens of millions of yuan annually in joint research.

Academic speakers discussed the VDT transformer‑based diffusion video generation model, data assetization strategies, and the need for comprehensive governance of AI‑generated video content, emphasizing standards development by the National Radio and Television Administration.

Looking forward, the Innovation Center will continue to focus on breakthrough technologies, strengthen industry‑academic‑research integration, and drive cross‑domain innovation to accelerate the digital and intelligent transformation of the broader video industry.