用 GPT Image 2.5 生成上百纸片元素做科普动画
一条给 Opus 5.5 的完整提示词：让它在 Fal AI 上用 GPT Image 2.5 批量生成剪纸风素材，做成 90–120 秒自进化 AI 智能体局限性的科普动态视频。
Hey Opus 5.5. Create a 90–120 second educational motion-graphics video in English about the current limitations of recursive self-evolving AI agents. Use a paper-cutout stop-motion style on a white background, with tiny square characters explaining large visualized concepts. Juicy motion design such as anticipation, overshoot, squash and stretch, and paper tear, flip and slide transitions. Build it from dozens or even hundreds of paper elements generated with GPT Image 2.5 on Fal AI. Before animating, study articles from the last three months on the limitations and open problems of self-evolving agents, and turn them into one engaging story. Keep all on-screen text plain, specific and grammatically correct, and avoid AI-slop patterns: no "not X, but Y" constructions, no three-fragment lists, no rhetorical questions, no dramatic ellipses