At the 2024 China Computer Conference, Zhang Di, Vice President of Kuaishou and Head of the Large Model Team, announced that KeLing AI will be conducting internal testing of its video face modeling feature. This feature, based on innovative ID preservation capabilities, allows users to self-train face models. Once trained, these models can be used to generate 5s/10s video content from text.
The aim of this feature is to address the challenge of creating stable character IPs in AI video generation. By self-training face models, users can create their own virtual characters and use these models to generate videos.
Since its launch in June this year, KeLing AI has amassed over 3.6 million users, who have collectively generated 37 million videos and over 100 million images. KeLing AI has made significant advancements and expanded applications in the field of video generation, including improvements in image quality, aesthetic expression, motion plausibility, and semantic understanding.
Zhang Di also revealed that KeLing AI will soon launch a standalone app for easier user access. Through the standalone app, users can more conveniently utilize KeLing AI's features, including the video face modeling function.