步骤 2：分析该视频很可能是由 GPT-image 2 加上 Seedance 2.0 生成的，并导出用于 GPT-image 2 和…

Step 2: Analyze that the video is likely generated with GPT-image 2 plus Seedance 2.0, and export the prompts for GPT-image 2 and Seedance 2.0! GPT- image 2 prompt 👇 Generate a 16:9 low-resolution early-2000s VHS home camcorder still frame. A heavy-set adult man with light skin, short dark beard, thin-rim glasses, wearing a black fedora and oversized plain black t-shirt, sits at a cluttered computer desk typing on a black keyboard. He is the main subject, positioned center-left, focused seriously on a large bulky beige CRT monitor in the left foreground. The monitor is close to camera and partially blocks the keyboard. The room is a cramped bedroom office with beige walls, a bright overexposed window with blinds on the far left, shelves packed with CDs, DVDs, game cases, anime boxes, and old media behind him. The desk is messy with CD cases, papers, a mouse, wrappers, and a blue soda can. Standing to the man’s right in the same room is a living anime cheerleader girl, naturally present beside him as if she is physically there. She has a clearly 2D anime-inspired face with large expressive eyes, brown hair, and stylized anime proportions, but she is integrated into the real VHS footage with matching lighting, shadows, lens blur, grain, and low-resolution camcorder artifacts. She wears a blue-and-white sailor cheerleader outfit and holds bright yellow pom-poms. No white outline, no cardboard border, no standee, no printed poster look. She should look like an anime character who has entered the real room, standing beside the man and cheering, slightly surreal but physically sharing the same space. Composition: 16:9 horizontal frame, CRT monitor large on the left foreground, man center-left, anime cheerleader girl on the right side behind and beside him, both visible in the same camcorder shot. The man and room feel realistic; the girl remains anime-stylized but blended into the footage. Authentic VHS home-video look: soft focus, interlaced scanlines, analog tape grain, chroma bleed, muted beige colors, low contrast, blown-out window highlights, noisy shadows, slight exposure flicker, mild barrel distortion, dirty tape edge artifacts, early digital compression. Awkward deadpan internet humor, mundane and surreal. No white outline, no cardboard cutout, no standee, no poster, no sticker border, no readable text, no watermark, no modern monitor, no cinematic lighting, no glossy clean AI look.

步骤 2：分析该视频很可能是由 GPT-image 2 加上 Seedance 2.0 生成的，并导出用于 GPT-image 2 和 Seedance 2.0 的提示词！ GPT-image 2 提示词 👇 生成一张 16:9 低分辨率的 2000 年代早期 VHS 家用摄像机静态画面。 一个体格壮硕的成年男子，肤色较浅，留着短深色胡须，戴着细框眼镜，头戴黑色软呢帽，身穿宽大的纯黑色 T 恤，坐在一张凌乱的电脑桌前敲打一个黑色键盘。他是画面主角，位于画面中偏左，神情专注地盯着一台位于左前景的体积庞大、米色的 CRT 显示器。显示器距离摄像机很近，部分遮挡了键盘。 房间是一个狭小的卧室办公室，米色墙壁，左侧远处有一扇带有百叶窗、明亮过曝的窗户，他身后是堆满 CD、DVD、游戏盒、动漫盒子和旧媒介的架子。桌面杂乱地散落着 CD 盒、纸张、鼠标、包装纸和一只蓝色汽水罐。 在该男子的右侧、同一个房间里，站着一个活生生的动漫啦啦队女孩，自然地出现在他身旁，仿佛她真的物理存在于那里。她有一张明显是二维动漫风格的脸，眼睛大而有神，棕色头发，身材比例是风格化的动漫比例，但她被融合进真实的 VHS 影像中，光照、阴影、镜头模糊、颗粒感和低分辨率摄像机伪影都相互匹配。她穿着蓝白配色的水手服啦啦队服装，手持亮黄色的啦啦队花球。不要白色描边，不要纸板边框，不要人形立牌，不要印刷海报的感觉。她应该看起来像一个动漫角色走进了真实房间，站在男子旁边加油鼓劲，带有一丝超现实感，但物理上共享同一个空间。 构图：16:9 横向画面，CRT 显示器在左前景占据较大位置，男子位于中偏左，动漫啦啦队女孩在他右后方和右侧，两人都出现在同一个摄像机镜头中。男子和房间感觉是写实的；女孩则保持动漫风格化，但融入到影像里。 真实的 VHS 家庭视频外观：柔焦、隔行扫描线、模拟磁带颗粒、色彩溢出（chroma bleed）、哑光米色色调、低对比度、过曝的窗户高光、噪点阴影、轻微曝光闪烁、轻微桶形畸变、脏污磁带边缘伪影、早期数字压缩感。尴尬的冷面互联网幽默，平淡又超现实。 不要白色描边，不要纸板人形立牌，不要立牌，不要海报，不要贴纸边框，不要可读文字，不要水印，不要现代显示器，不要电影级打光，不要光滑洁净的 AI 观感。