Generative video continues to be one of the fastest-advancing areas in the AI field, and at this year’s CES, Kling AI made a strong case that it’s more than just an experimental playground. The company unveiled key user growth milestones while demonstrating the capabilities of a new model that represents a more complete end-to-end AI video production workflow.
One of the biggest draws for Kling AI these days is its motion control capabilities, powered by the latest Video 2.6 model. The feature has recently gone viral on social media, allowing users to combine a single photo with a reference motion clip (such as a dance or facial expression) to generate a short AI video in about a minute. The results were widely shared online, and trends like “dancing puppies” propelled the app to the number one download spot in multiple markets.
This surge in popularity has made Kling AI the most downloaded app in four countries, including South Korea and Turkey, and in the top 10 in 10 other markets, according to a JPMorgan report. The same motion control functionality was demonstrated live at Kling AI’s CES booth, where attendees could see how quickly static images can be transformed into expressive animated clips.
A unified approach to AI video creation
Beyond its viral capabilities, Kling AI used CES to highlight its long-term vision for AI video creation. The company showcased the O1 model, which it describes as the industry’s first integrated multimodal video model that combines generation, editing, and understanding into one system.
One of the most notable features of the O1 model is prompt-based post-production editing. Instead of relying on traditional editing software, users can simply enter instructions such as “eliminate bystanders,” “change daytime to dusk,” or “swap the main character’s outfit.” The model understands the visual context of the video and applies these changes directly, covering everything from background elements and lighting to weather, clothing, and time of day.
Another key highlight is multi-subject consistency, a long-standing challenge in AI video generation. O1 models are designed to maintain consistent characters, props, and environments throughout a scene, even when the camera moves or multiple subjects interact. Kling AI positions this feature as close to how a human director would oversee continuity throughout a shoot.
The company also demonstrated the native audiovisual generation of the Video 2.6 model, which supports dialog, ambient sounds, and sound effects that are generated along with the video. This approach aims to reduce the need for separate tools and provide creators with a more streamlined end-to-end workflow.
Rapidly increasing user and enterprise adoption
Alongside the technology demo, Kling AI unveiled the latest platform metrics to show how quickly it is scaling. By the end of 2025, the platform has grown to enable the creation of over 600 million videos and has over 60 million users worldwide. Today, over 30,000 companies and developers are integrating Kling AI’s APIs into their products and workflows. This is topped by December 2025 revenue of US$20 million, and the company’s annual operating rate is over US$240 million.
This represents a significant increase from numbers released just a few months ago. In July, the company reported 45 million users and 20,000 business customers. Today, the company’s enterprise user base spans industries such as advertising, animation, gaming, and film production, with notable clients such as Higgsfield, ComfyUI, Fal.ai, and Freepik.
AI movie production exhibition
Kling AI also focused on creative storytelling at CES, sharing an AI-powered short film, “A Very AI Yule Log,” with attendees. Directed by Jason Zada, the project was created in collaboration with Secret Level and used generative AI to reimagine the classic holiday fireplace video.
The movie has more than 600 AI-generated scenes, rather than static loops, each about 10 seconds long, adding up to nearly two hours of continuous content. Surreal elements appear and disappear around the fireplace. All are generated between fixed start and end frames.
Speaking on the Kling AI panel titled “How GenAI is transforming the creative industries,” Zada explained how quickly the project came together. He noted that a similar idea attempted a year ago felt premature, but advances in Kling AI’s model now allow it to produce nearly two hours of original visuals and AI-generated music within two weeks.
From viral clips to production infrastructure
Taken together, Kling AI’s CES exhibit highlighted how generated video is progressing beyond novelty. Viral social features may be driving downloads, but under the hood, the platform is evolving into a complete production toolset for creators and businesses alike.
As AI video models continue to mature, Kling AI’s combination of rapid user growth, integrated editing and generation, and real-world creative projects suggests this technology is becoming more permanent. CES felt less like a glimpse of the future and more like the early stages of new production standards taking shape.
