PROMPT RADAR / VIDEO视频叙事
gpt 5.6 + seedance 2.0 should honestly be illegal posted a side by side
gpt 5.6 + seedance 2.0 should honestly be illegal posted a side by side comparison yesterday and half of the replies got it wrong so here's round 2: one of these girls is real. the other is the ai rebuild. same room this
@twoclipping 2.2k 热度 6.5 万 浏览
编辑先看
它为什么值得进入灵感库
这条内容值得留下的，不是表面的风格词，而是它提供了一条可以继续实验的画面路径。 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。
01 / 原始方法
动作与镜头 Prompt
gpt 5.6 + seedance 2.0 should honestly be illegal posted a side by side comparison yesterday and half of the replies got it wrong so here's round 2: one of these girls is real. the other is the ai rebuild. same room this time, no environment tells the receipts: $400 in creator fees for the real version. the ai one costs about $2 inside the system im running: - the character ref prompts (named imperfections stack) - the video prompts + the anti ai negatives - the one line fix every gpt 5.6 script needs - the full rebuild workflow comment LEFT or RIGHT and i'll dm you the full prompting system. must be following so the dm goes through
02 / 复刻路径
不要一次改完所有变量
用 Seedance 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。
- 主体：gpt 5.6 + seedance 2.0 should honestly be illegal posted a side by side comparison yesterday and half of the replies got it wrong so here's round 2: one of these girls is real. the other is the ai rebuild. same room this
- 视觉方向：实时 AI 视频案例；查看原帖媒体与作者说明
- 镜头与构图：按原帖镜头流程执行；查看原帖媒体与工作流
- 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
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