A squirrel in New York’s Central Park tried to drink the author’s iced coffee, and that photo became an experiment. Within 21 minutes, LibTV Agent, a new AI video platform from Chinese company Liblib, turned it into a 30-second Pixar-style animation. The process revealed something far more significant than speed: the Agent replicated the entire workflow of a professional director, from character asset creation to storyboarding and editing. It marks a shift from "one-shot generation" to a controllable, iterative production pipeline.
The author, sitting on a park bench, uploaded the photo to the Agent and selected a Skill labeled "Pixar Animation Ad." A Skill is essentially a job description for AI, containing the steps and principles that a real director would follow for that type of piece. The Agent recognized the squirrel, the green straw, the iced drink, and the beige tote bag. It asked only one question: landscape or portrait orientation. Then it proposed four character designs, and the one chosen featured the squirrel in tiny sunglasses, looking like a street-smart New Yorker. This moment highlighted a critical innovation: the Agent didn’t just generate a video blindly—it stopped for human approval at key checkpoints.
What truly stood out was the "anchor point" system. Before any animation began, the Agent created separate "asset portraits" for each element—the coffee cup, the bag, the bench—and paused to let the user confirm every detail. This ensured consistency across all 18 shots that followed. The final output even retained the Starbucks logo with near-advert quality. The real magic isn’t that AI can make a 30-second video—it’s that you can say ‘no’ at every step. The author noted that this level of control is rare among AI video tools; most competitors like Runway Gen-3 or Pika generate fluid snippets but fail to maintain character identity across scenes, leading to the dreaded "morphing face" problem.
Next came the storyboard. The Agent produced a grid of pencil-like sketches, each frame labeled with camera movements—"Extreme close-up, Dutch angle 15 degrees." The author recalled seeing similar storyboard walls at the Academy Museum in Los Angeles, where Hitchcock’s team believed the film was half-done once the storyboard was finished. Seeing an AI draw that same document felt like taking a masterclass in cinematography on the fly. The two parallel video assets were generated simultaneously, accompanied by a jazz BGM, and automatically assembled in a built-in timeline editor with dual tracks, subtitles, and transitions. No external editing suite was needed.
Beyond the squirrel story, the author tested two more briefs—a product advertisement and a short brand film—to gauge consistency. In each case, the anchor-point mechanism prevented asset drift, and the Skill library offered tailored workflows (e.g., "Product Highlight" and "Corporate Story"). This reveals what LibTV Agent is really selling: not just video generation, but a marketplace of reusable filmmaking expertise. Skill Store is the app store of creative AI—it packages a professional’s craft into a downloadable skill. Some critics argue that such automation strips away the human touch, but the author found the opposite: the Agent demands active decision-making at every stage, from choosing character style to approving camera angles. It empowers non-professionals while still requiring a director’s vision.
The platform itself is no newcomer. Launched in March 2024, LibTV attracted over 100,000 creators on its first day. The team version, covered in an earlier article, already supported collaborative workflows. The Agent release completes a trilogy: first the canvas, then the team, now the AI co-director. For context, China’s AI video market is heating up, with players like Kling and Vidu also releasing text-to-video models, but few address the production workflow holistically. LibTV Agent fills that gap by keeping the human in the loop, turning AI from a black box into a transparent co-worker.
The most profound takeaway from the 21-minute squirrel film is this: the bottleneck of AI creativity has shifted from generating content to directing it. The Agent removed the need for technical rendering skills, but it amplified the need for creative judgment. Users who lack a sense of story or pacing will still get mediocre results—the AI faithfully follows, but it does not innovate alone. This aligns with a broader industry trend: AI tools are becoming less autonomous and more assistive, emphasizing control and iteration over pure speed.
In the end, the author exported the 30-second clip with a click. The squirrel’s eyes darted, the straw bent, the jazz score swelled. It wasn’t perfect—some frames had slight jitter, and the animation style leaned generic rather than truly Pixar-grade. But the process was transparent, adaptable, and repeatable. For anyone who has felt intimidated by video creation, LibTV Agent offers a way in: speak your goal, approve each step, and learn the craft as you go. The director’s manual is now open to anyone, one anchor point at a time.