Kuaishou: Kuaishou Technology Unveils " Multi-Image Reference" Feature to Further Tackle AI Video Consistency Challenges

Kuaishou Technology recently announced the global launch of Kling AI's " multi-image reference" feature, addressing one of the most persistent challenges in AI video generation - maintaining visual consistency. This cutting-edge technology enables AI to analyze and integrate diverse subjects from multiple images and creatively generate composite videos based on users' text descriptions, empowering creators to produce seamless and professional-quality videos with greater precision. The " multi-image reference' feature is now available to users worldwide through the Kling AI 1.6 model. Users can upload one or more images of the same subject (such as a person, animal, movie character or object), and the model will reference these images to generate a video with a consistent style, ensuring the subject remains visually consistent across multiple scenes. For example, by uploading images of a cat, a jacket, and sunglasses and entering a prompt like "A humanoid cat in a jacket and sunglasses turns on stage and poses for the camera," the model skillfully blends the reference images and user instructions to produce a professional-grade video. Users can also use reference images to define specific scenes, enabling characters to don different outfits and undertake actions within designated settings, thus greatly enhancing creative control. For example, by uploading reference images of an elderly man, a cup of coffee, and a cafe scene, and entering a prompt such as "A cartoon-style elderly man grabs a cup of coffee in a cafe, users can generate a video depicting a man sitting in a cafe sipping his coffee. Furthermore, the " multi-image reference" features enables dynamic interactions between different characters. Users can upload images of multiple subjects (such as people, animals or movie characters) and describe their desired interactions through text prompts. For example, by uploading Reference images of a boy and a Corgi and entering the prompt "A cute boy is petting a Corgi," the feature will generate a video that brings the scene to life (video frame shown below). Since its initial rollout on June 6, 2024, Kling AI has undergone dozens of extensive iterations and upgrades. Most notably, in December 2024, the video generation model Kling AI 1.6 was officially launched, building upon the foundation model and elevating its capabilities significantly. Kling AI 1.6 introduced major advancements in text responsiveness, increasing its ability to interpret text descriptions related to motion, temporal action, and camera movement. Additionally, the upgraded model enhanced motion quality with smoother motion and more natural expressions. Kling AI 1.6 also delivered comprehensive upgrades in color accuracy, lighting dynamics, and detailed rendering, resulting in a striking improvement in overall video quality. Users can access the new feature via the web portal (Chinese version: or by searching for and downloading KLINGAI in the app store. With the launch of the " multi-image reference' features, Kling AI has taken a substantial step forward in addressing the consistency challenges often faced in video generation. This innovative feature empowers users to unleash their creativity by freely combining various visual elements to craft unique and imaginative scenarios, opening the door to endless possibilities for storytelling and content creation. Kling AI now serves a growing user base of over 6 million users, with more than 65 million videos and 175 million images generated to date. The platform is also expanding its monetization efforts through tailored membership services for individual users and API interface solutions for businesses, where it has partnered with thousands of enterprise customers, accelerating its monetization progress. Kuaishou Technology recently announced the global launch of Kling AI's " multi-image reference" feature, addressing one of the most persistent challenges in AI video generation - maintaining visual consistency. This cutting-edge technology enables AI to analyze and integrate diverse subjects from multiple images and creatively generate composite videos based on users' text descriptions, empowering creators to produce seamless and professional-quality videos with greater precision. The " multi-image reference' feature is now available to users worldwide through the Kling AI 1.6 model. Users can upload one or more images of the same subject (such as a person, animal, movie character or object), and the model will reference these images to generate a video with a consistent style, ensuring the subject remains visually consistent across multiple scenes. For example, by uploading images of a cat, a jacket, and sunglasses and entering a prompt like "A humanoid cat in a jacket and sunglasses turns on stage and poses for the camera," the model skillfully blends the reference images and user instructions to produce a professional-grade video. Users can also use reference images to define specific scenes, enabling characters to don different outfits and undertake actions within designated settings, thus greatly enhancing creative control. For example, by uploading reference images of an elderly man, a cup of coffee, and a cafe scene, and entering a prompt such as "A cartoon-style elderly man grabs a cup of coffee in a cafe, users can generate a video depicting a man sitting in a cafe sipping his coffee. Furthermore, the " multi-image reference" features enables dynamic interactions between different characters. Users can upload images of multiple subjects (such as people, animals or movie characters) and describe their desired interactions through text prompts. For example, by uploading Reference images of a boy and a Corgi and entering the prompt "A cute boy is petting a Corgi," the feature will generate a video that brings the scene to life (video frame shown below). Since its initial rollout on June 6, 2024, Kling AI has undergone dozens of extensive iterations and upgrades. Most notably, in December 2024, the video generation model Kling AI 1.6 was officially launched, building upon the foundation model and elevating its capabilities significantly. Kling AI 1.6 introduced major advancements in text responsiveness, increasing its ability to interpret text descriptions related to motion, temporal action, and camera movement. Additionally, the upgraded model enhanced motion quality with smoother motion and more natural expressions. Kling AI 1.6 also delivered comprehensive upgrades in color accuracy, lighting dynamics, and detailed rendering, resulting in a striking improvement in overall video quality. Users can access the new feature via the web portal (Chinese version: or by searching for and downloading KLINGAI in the app store. With the launch of the " multi-image reference' features, Kling AI has taken a substantial step forward in addressing the consistency challenges often faced in video generation. This innovative feature empowers users to unleash their creativity by freely combining various visual elements to craft unique and imaginative scenarios, opening the door to endless possibilities for storytelling and content creation. Kling AI now serves a growing user base of over 6 million users, with more than 65 million videos and 175 million images generated to date. The platform is also expanding its monetization efforts through tailored membership services for individual users and API interface solutions for businesses, where it has partnered with thousands of enterprise customers, accelerating its monetization progress.