Uisato Studio 正在悄然重新思考 AI 视频的工作方式——而且它并不是你所预期的那样。

Uisato Studio is quietly rethinking how AI video works — and it's not what you expect. Most AI video tools are prompt boxes with a generate button. Type something, wait, get a clip, repeat. Uisato treats it like a film production pipeline: 1. Performance capture first — body language, movement, choreography 2. Style direction second — visual aesthetic, color grading, lighting 3. Generation follows the performance, not the other way around This is the opposite of what Runway, Pika, and Sora do. Instead of prompting for "a woman walking down a street," you capture the walk first, then tell the AI what style to render it in. The demo clip "There There" (ft. Sara Silkin) shows the difference. The motion feels intentional because it IS intentional — a human performed it, the AI just painted over it. This is where AI video is heading: motion capture as input, generation as render pass. Not prompt → clip, but perform → direct → render. Uisato Studio: https://t.co/pUujA5d7RI

Uisato Studio 正在悄然重新思考 AI 视频的工作方式——而且它并不是你所预期的那样。 大多数 AI 视频工具就是一个带生成按钮的提示框。输入一些内容，等待，得到一个片段，重复。 Uisato 将其视为电影制作流水线： 1. 首先是表演捕捉——肢体语言、动作、编舞 2. 其次是风格指导——视觉美学、调色、布光 3. 生成跟随表演，而不是反过来 这与 Runway、Pika 和 Sora 所做的完全相反。你不是输入提示比如「一个女人走在街上」，而是先捕捉走路的动作，然后告诉 AI 以什么风格来渲染它。 演示片段《There There》（嘉宾：Sara Silkin）展示了这种差异。动作感觉是有意为之，因为它确实是有意为之——由真人表演，AI 只是在其上进行绘制。 这就是 AI 视频的发展方向：以动捕作为输入，生成作为渲染通道。不是 prompt → clip，而是 perform → direct → render。 Uisato Studio: https://t.co/pUujA5d7RI