GPT Image 2.5 加视频模型做伪 DV 短片
用 GPT Image 2.5 配 Sedance 2.5 生成 30 秒首尔咖啡馆日常片段，提示词强调 2000 年代 DV 手持质感。
Made with GPT image 2.5+ Sedance 2.5 on @ImagineArt_X Prompt 👇 Create a 30-second ultra-realistic personal home-video of a young Korean woman visiting a small indie café in an older Seoul neighborhood on a quiet afternoon. Use the attached image as the character reference and keep her face, hairstyle, and outfit consistent throughout. She walks into a cozy, small café, orders at the counter, and waits briefly before receiving her coffee. She finds a seat by the window, pulls out a book from her tote bag, and settles in, sipping her coffee occasionally while reading. She pauses to look out the window for a moment, watching the street outside, then goes back to her book, turning a page and adjusting her position in the chair. Use raw early-2000s consumer DV-camera footage: handheld shake, imperfect framing, autofocus hunting, exposure shifts from window light, soft detail, mild noise, natural motion blur and occasional awkward zooms. Natural café ambience only — quiet murmur of other customers, cups clinking, page turns, faint street sounds through the window. No music, no narration, no dramatic events, no polished commercial cinematography.