PROMPT RADAR / VIDEO视频叙事

Midjourney v8.2 + Seedance 2.0 Mini --sref 74117661 --stylize 1000 --pro

Midjourney v8.2 + Seedance 2.0 Mini --sref 74117661 --stylize 1000 --profile ycjrvus --preview Prompt: The swordswoman @[reference image] launches forward from complete stillness, crossing the space in a single compresse

@aimikoda 937 热度 1.2 万 浏览
编辑先看

它为什么值得进入灵感库

它最有价值的部分,是把一个模糊想法推进到了可以被模型执行的程度。 把一段描述拆成起始状态、动作变化、镜头响应和结束画面,连续性比形容词更重要。

动作与镜头 Prompt

Midjourney v8.2 + Seedance 2.0 Mini --sref 74117661 --stylize 1000 --profile ycjrvus --preview Prompt: The swordswoman @[reference image] launches forward from complete stillness, crossing the space in a single compressed burst as blue and violet trails tear behind her body. Her blade cuts through the frame, triggering a circular shock explosion filled with orange pigment fragments, followed by several delayed color trails that reveal the full path of the strike. 16 cuts. No background music or score. High-end cinematic 3D realism fused with graphic painterly anime action. Hyper-fast sword choreography, compressed acceleration, chromatic motion trails, explosive shockwave bursts and fractured pigment splashes define every movement. Each strike releases expanding pressure rings, directional color ruptures and sharp ribbon-like trails that remain suspended for a split second before breaking into glowing fragments. Aggressive tracking shots, extreme perspective, rapid orbiting angles and sudden impact close-ups amplify speed while deep negative space, low-key lighting and controlled blue, violet and orange accents preserve strong silhouette readability. Realistic body weight, metallic materials and feature-film rendering ground the action beneath the abstract color physics.
查看原作者内容

不要一次改完所有变量

用 Midjourney 复刻时,建议先单独生成稳定首帧,再把动作拆成“开始、变化、收束”三段;每段只保留一个主要运动。

  1. 主体:Midjourney v8.2 + Seedance 2.0 Mini --sref 74117661 --stylize 1000 --profile ycjrvus --preview Prompt: The swordswoman @[reference image] launches forward from complete stillness, crossing the space in a single compresse
  2. 视觉方向:实时 AI 视频案例;查看原帖媒体与作者说明
  3. 镜头与构图:按原帖镜头流程执行;查看原帖媒体与工作流
  4. 交付约束:auto,保留原作者方法与归属;复刻时避免水印、低清、畸变、主体漂移和无关文字。
负向约束

保留原作者方法与归属;复刻时避免水印、低清、畸变、主体漂移和无关文字。

模型建议

Midjourney / Seedance / Veo / Kling。先用原始 Prompt 建立基准结果,再根据模型对自然语言、镜头运动或文字排版的能力做局部改写。

来源与编辑原则

保留作者,也保留判断。

本页收录公开案例并提供中文索引、结构化拆解和复刻建议,不冒充原作者作品,也不替代原帖上下文。

作者:Kōda · 收录于 2026年7月13日

访问原始来源

沿着这条线继续看