地铁乘客云团变形

固定机位拍摄对称地铁车厢，乘客头部在真人与不同天气云团之间进行数字故障切换，并在第 8 秒收束到参考图构图；原帖同时公开了完整画面与声音 Prompt。

生成结果

多参考生成

输入参考素材

0 / 1 个文件已公开

按提示词中的编号使用素材；输入参考与上方生成结果分开展示。

来源未公开全部参考文件；复现时请为缺失输入准备本人或已获授权的素材。

Picture 1 · 来源未提供文件

复现时，仅使用本人或已获授权的肖像、声音及其他素材。

提示词信息
分享者	@AION
模型	MiniMax H3
分类	参考生成与编辑
数据集	SkyNotSilent · awesome-MiniMax-H3-cases

完整提示词

integrated_multimodal_description: [Shot 1] Surreal live-action cinematic imagery in a symmetrical square deep-focus wide shot inside the same modern subway carriage established by <Picture 1>. At the beginning, the same seated passengers occupy the exact same seats with the same bodies, clothing, poses, hands, phones, bags, spacing, and calm commuter behavior, but every passenger has a distinct, anatomically normal human head and natural human face; the ceiling above them is ordinary dim stormy condensation without the giant eye. The camera holds a completely static shot for the entire video. From 00:01.600 to 00:02.700, a thin horizontal digital phase glitch sweeps across the carriage with brief RGB channel separation and blocky temporal displacement restricted to the passengers' heads; their normal human heads transform into compact volumetric storm clouds while their necks, bodies, clothing, hands, and positions remain perfectly unchanged. From 00:02.700 to 00:03.450, the interference reverses and restores the same normal, stable human faces. From 00:03.450 to 00:05.550, two faster synchronized glitch pulses alternate every head between the same normal human face and its cloud form, with no body movement, no face melting, and no change to the carriage geometry. From 00:05.550 onward, a final sustained digital stutter locks the passengers into their distinct cloud-headed states: lightning flickers inside the left foreground cloud, warm light glows inside the left midground cloud, rain falls beneath the small dark cloud, a faint rainbow forms in the right-center cloud, and a small sun glows inside the right foreground cloud. Fine luminous mist rises from the central cloud and resolves into the huge pale cloud face and single open eye overhead. The glitches settle, all motion becomes nearly still, and at exactly 8.00 seconds the passenger count, cloud shapes, body poses, wardrobe, seats, poles, windows, lighting, colors, crop, ceiling apparition, and full composition converge precisely to <Picture 1>. No cuts, no camera movement, no zoom, no pan, no tilt, no roll, no shake, no body morphing, no extra people, no missing passengers, no fused limbs, no wardrobe changes, no seat changes, no floating bodies, no text, no subtitles, no logos, and no drift from the final reference image.

overall_soundscape: A steady low subway rumble, muted wheel rhythm, and soft ventilation hum continue throughout. Each transformation is marked by a brief dry digital stutter and electrical crackle, while the final cloud state adds faint rain, distant thunder, and a soft airy hiss without overpowering the train ambience.

non_diegetic_music: N/A

在 DraPub 中尝试

看看同一视觉方向里的其他提示词。

返回提示词画廊

从简化网格到角色拳击动作

参考生成与编辑

从简化网格到角色拳击动作

左右画面同步展示简化人体网格与原角色骨骼，拳、拳、踢的组合动作被生成后重定向回完整角色。原生视频规格为 12 秒 · 1040×588。原帖完整公开了 Prompt，本页按原文保留。

双手从零搭起一座体育场

参考生成与编辑

双手从零搭起一座体育场

一双手在空白工作台上逐步铺草皮、搭看台并补齐座椅，把微缩体育场完整组装出来。原生视频规格为 15 秒 · 2048×1152。原帖完整公开了 Prompt，本页按原文保留。

漫画页面逐格变成动态图解

参考生成与编辑

漫画页面逐格变成动态图解

原始漫画页面与局部放大动画并列，依次突出人物、文字和操作步骤。

分镜板指导动漫料理过程

参考生成与编辑

分镜板指导动漫料理过程

料理动画与分镜板并排展示，呈现参考面板到连续动作的对应。