PROMPT RADAR / VIDEO视频叙事

ChatGPT + Seedance 2.0 Workflow Step 1: Create a new ChatGPT project and

ChatGPT + Seedance 2.0 Workflow Step 1: Create a new ChatGPT project and upload everything you want the model to remember throughout the production, including character sheets, location references, style guides, product

@beginnersblog1 501 热度 5.3k 浏览
编辑先看

它为什么值得进入灵感库

先别急着照抄整段 Prompt:真正有用的是主体、空间和镜头之间的约束关系。 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

ChatGPT + Seedance 2.0 Workflow Step 1: Create a new ChatGPT project and upload everything you want the model to remember throughout the production, including character sheets, location references, style guides, product references, story notes, and continuity rules. Step 2: Define the objective. Describe the story, commercial, trailer, or scene you want to create and explain the visual style, pacing, and overall creative direction. Step 3: Ask ChatGPT to break the story into multiple 5 to 10 second batches. Each batch should have a clearly defined start frame and end frame. The end frame of one batch must become the exact start frame of the next to maintain perfect continuity. Step 4: Ask ChatGPT to generate a detailed storyboard for every batch. Keep the same characters, costumes, props, lighting, scale, camera direction, and environment consistent across all frames. Step 5: Review the storyboard. Regenerate individual batches until every transition, action, and composition feels natural and continuous. Step 6: Once the storyboard is approved, ask ChatGPT to convert each batch into a professional Seedance 2.0 video prompt. Each prompt should include cinematography, camera movement, character performance, continuity rules, environment details, and animation instructions. Step 7: Generate every batch in Seedance 2.0. Render them in parallel for speed or one by one if you want to review each scene before continuing. Step 8: If any shot breaks continuity, regenerate only that specific batch while keeping the same start frame, end frame, characters, and visual style. Step 9: Stitch all generated clips together in chronological order. Since every batch shares matching transition frames, the complete video should play as one seamless, continuous sequence. Step 10: Perform the final edit by adding sound effects, music, dialogue, color grading, and any finishing touches before exporting the completed video.
查看原作者内容

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。

  1. 主体:ChatGPT + Seedance 2.0 Workflow Step 1: Create a new ChatGPT project and upload everything you want the model to remember throughout the production, including character sheets, location references, style guides, product
  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

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:Beginnersblog · 收录于 2026年8月8日

访问原始来源

沿着这条线继续看