The multiple image reference student video upgrade of the Keling AI has increased the model's efficiency by 102%.

On July 24th, the multi-image reference model and function of KeLing AI video were upgraded significantly. After blind testing, the new model's performance has improved by a whopping 102% compared to the previous one. With the upgrade, the videos generated through the multi-image reference function have more consistent character roles, significantly improved dynamic quality, more reasonable and natural interactions between multiple subjects, and smoother motion performance. Additionally, the overall image quality has been greatly enhanced, resulting in a more refined overall effect. Furthermore, the upgrade to the multi-image reference function now supports users to only reference specific parts or areas of images, such as specific subjects, facial features, or clothing, to allow the model to more accurately reference and avoid the interference of irrelevant elements, making it easier to achieve a perfect result.