成片拆解Ima Studio 官方示例:MiniMax H3 生成 K-fashion 图形拼贴 MV,一个女孩 8 套造型。
结构:封面卡闪现 → 8 套造型约 1.6 秒一切 → 杂志封面定格
看完成片再对照:这条片子靠什么在几秒内抓住人。点时间可跳到成片对应位置。
第 0 秒先闪约 0.5 秒的杂志封面卡:女孩伸手指向镜头,巨大斜体「TRENDY LOOK」压在画面中间,四周是 K-POP STYLE、STAY COOL 等贴纸字。
约 0.5s「SEOUL POP」条纹背景,约 2.1s「SOFT CHIC」粉蓝波点色块全身,约 3.7s 黑白棋盘格坐姿托腮,约 5.3s 紫底「LOVELY MOOD」竖排字——大约每 1.6 秒卡点切一套版式。
约 6.9s 手掌伸向镜头配「TRENDY LOOK」,约 8.5s 米白底脸部特写,约 10.2s 黑色弧线图形前的全身站姿,约 11.8s 菱形棋盘格前手扶耳侧。
约 13.5s 前面的棋盘格、波点、条纹圆和「SEOUL POP」大字重新拼成一张杂志封面,她站在中间定格到结束。
提示词先锁死同一个人,再把图案、配色、服装、动作、字体、时间线分块写清;动作用「半步向前、回头、撩头发、伸手向镜头」这种 pose-to-pose 清单,结尾要求所有图形重新拼成封面。
别急着上视频模型。这个片子的秘密是:先把角色和场景锁死,再一次生成 10 个镜头。
提示词第一句写明基于「attached single character reference image」(一张人物参考图)生成,整条都要锁定同一个韩系女孩的脸、发型和身材。这张参考图原帖没有公开,需要自己准备一张。原帖没有故事板,也没有驱动视频。
这是 Ima Studio 官方账号发的产品示例帖,主帖写用的是 MiniMax H3,并附了自家 AI 音乐 MV 生成器的推广链接(本页不收录)。提示词要求 15 秒、16:9、24fps;X 上的成片是 3840×2160、30fps、约 15.7 秒。画面里没看到水印。声音与成片差异:成片从头到尾是一段节奏很满的配乐,剪辑基本卡在节拍上;没有识别到台词或旁白,也没有单独的音效。提示词里完全没写声音,配乐来源原帖没说明(Ima Studio 这个工具的卖点是上传音频生成卡点 MV,但帖子没明说这条是不是这样做的)。与提示词不符:开头多了约 0.5 秒的贴纸风封面卡;结尾封面上有一组字母变形、读不出来(约像「JOEFV COOD」),而提示词要求文字清晰可读。
把下方英文提示词整段粘贴。分块依次是:Character lock(锁人)→ Style → Visual system(图案清单)→ Color → Fashion → Motion & camera(pose-to-pose 动作清单)→ Typography(可用关键词)→ Timeline(8 套造型 + 封面定格 0.5 秒)→ Avoid。
方便你对照成片理解每一镜在干什么。
直接全选复制到 Seedance。@Image 1–0 必须对应上面 REF 编号。
MiniMax H3 · 主帖英文完整提示词(开头宣传语、推广链接和话题标签未收录)
Create a 15-second, 16:9, 24fps high-quality fashion MV based on the attached single character reference image. Character lock: Keep 1 young Korean-style woman only throughout the entire video. Strictly preserve her face, facial proportions, skin tone, age, long dark-brown hair, bangs, body proportions, limb proportions, identity, and overall Korean beauty vibe. She must always remain the same person. No face drift, no westernized features, no masculine look, no second person, no crowd. Style: Pattern Collage Pop × Korean Fashion Editorial × Fashion MV. The feeling is a stylish Korean fashion girl entering a magazine world made of classic patterns, geometric graphics, collage layouts, and bold typography. Keep the girl realistic, but make the background and design elements flat, sharp, clean 2D graphic design. Avoid rough handmade collage; make it polished, chic, trendy, glossy, and editorial. Visual system: Use patterns and shapes such as stripes, polka dots, checkerboard, plaid, grids, zigzags, borders, concentric circles, triangles, circles, squares, geometric lines, graphic frames, cutout color blocks, paper-edge collage shapes, overlapping color fields, whitespace, and offset layouts. Use them as backgrounds, framing devices, cross-screen bands, layered blocks behind the subject, and magazine-style title areas. Color: Extract 2 main colors from the girl’s vibe, plus white or black as support. Keep the palette clean and unified. Preferred tones: cream white, light blue, icy blue, silver gray, soft pink, muted lavender. Fashion: Extend the styling from the reference into sweet-cool Korean / light Y2K / chic Asian fashion. Possible items: light denim, fitted tops, cute-cool mini skirts, small bags, necklaces, earrings, platform ankle boots, Korean accessories. Slight outfit variation is allowed, but she must still feel like one continuous character. Motion & camera: This is a fashion MV, not a slideshow. Each segment must have a clear pose-to-pose action: half-step forward, look back, lower gaze then look up, fix hair, touch bag strap, hand near face, seated side look, crouch with soft smile, reach toward camera, light turn, final hero pose. Expressions: soft, sweet-cool, confident, polished. Use fast but smooth lateral movement, light push-in/pull-out, alternating close/medium/full shots, detail inserts, and editorial split compositions. Typography: Add large, readable fashion typography without covering the face. Use a main NAME title, plus supporting English keywords and occasional Korean text. Words can include: SEOUL POP, CUTE SIGNAL, SOFT CHIC, KOREAN GIRL, FASHION CUT, LOVELY MOOD, TRENDY LOOK. Vary placement, size, direction, and color while keeping a unified magazine design system. Timeline: 8 stylish looks across 15s, ending in a strong final hero shot where previous graphics and keywords reassemble into a polished Korean magazine-cover composition. Hold the final frame for 0.5s. Avoid: multiple people, identity drift, static poster feel, stiff motion, overly realistic backgrounds, unreadable text, text over face, childish tone, cheap filters, logos, watermarks, subtitles, exaggerated dance, or goofy expressions.
当前是单页静态站:视频 + 指引 + 可复制提示词。后面可以做成多教程列表、账号、收藏提示词、一键复制、手机竖屏版等。