时长与镜头：原作按 30 秒设计、每 2–3 秒一个时代。Seedance 2.0 官方说明单条最长 15 秒，九个时代塞不进去，实际用时建议只选 4–5 个，例如"远古 → 秦汉 → 盛唐 → 近代 → 当代"；想要完整版就分两条：第二条用第一条的最后一帧作首帧继续跑。

最值得学的是"遮挡转场"：让一个真实物体（车轮、旗子、门、蒸汽）贴着镜头扫过 0.2–0.8 秒，在被挡住的那一瞬间完成换装换景。观众会觉得是一镜到底，模型也有了"合理换画面"的机会，比直接要求"背景变成唐朝"稳得多。这个方法同样适合做"一镜到底换装""一镜到底四季变化"。

怎么用：@图片1 用你自己的正脸清晰照片，并在提示词里强调"只锁身份、不继承服装背景"。时代列表里的 [变量] 可以改成任何时间线，比如"一座城市的 1950 → 1980 → 2000 → 现在"，或者"一个人的童年 → 少年 → 中年"。

常见失败与调整：

• 每个时代换了一张脸：减少时代数量，并把"脸部严格以 @图片1 为准"放在提示词最前面。
• 出现硬切：把转场物体写得更具体（"一扇厚重的木门从镜头前滑过，完全遮住画面半秒"）。
• 古代场景像影视城：加"真实的泥土、灰尘、磨损的器物、忙碌的普通人"。

改编自 @QtImVCr6WK56152 发布、YouMind-OpenLab/awesome-seedance-2-prompts 收录的提示词，许可证 CC BY 4.0。

英文原版（原文较长，此处节选）（原帖为中文，以下为仓库收录的英文译本）

Generate a [30-second, 16:9 horizontal, 4K, live-action, high-budget cinematic quality] "Chinese Civilization Time Travel" short film, deeply referencing the actual camera structure and time progression logic of my uploaded reference video: This is not an edit of multiple historical scenes, but a single camera following [the same Chinese male] continuously forward from second 0 to 30, traveling from ancient times through thousands of years of Chinese civilization, and finally continuing past contemporary times into the future like the reference video; Use my uploaded male photo as the [sole and absolute face reference] for the entire film, the photo is only responsible for locking the person's identity, strictly maintaining his face shape, feature proportions, eyebrows/eyes, eye distance, nose bridge/tip, lips, jawbone, facial bone structure, skin tone base, youthful age sense, and recognizability. Absolutely do not reference the blue suit, tie, restaurant, chandelier, roses, jewelry, sitting posture, or background in the photo. Throughout 0-30s, viewers must clearly recognize this is the man from the reference image; prohibit changing actors per era, prohibit feature drift, prohibit becoming different ancient handsome guys or AI template faces. Era changes can only alter his hairstyle, hair length, binding method, beard, clothing materials, footwear, headwear, belts, tools, accessories, status, and habits. The entire film must visually be a true "one-take shot," prohibiting hard cuts, black screens, flash whites, fades, dissolves, sudden scene changes, character teleportation, or background refreshes. The character must never stand still letting the world change around him; he must actively move forward, running, walking fast, crossing, jumping down stairs, dodging sideways, grabbing vehicles, passing through crowds, pushing doors, climbing bridges, passing buildings, going past transport. The camera moves with him, naturally changing framing between side-front tracking, frontal backward tracking, lateral panning, rear chasing, low-angle footsteps, wide shots, and close-ups like the reference video. All era transitions should prioritize hidden cuts using real physical motion, e.g., rocks, pottery wheels, cart wheels, chariots, pillars, flags, fabrics, scrolls, doors, bridge columns, carriages, steam, trains, buses, crowds, glass curtains, transparent barriers passing extremely close to the camera and briefly obscuring for 0.2-0.8 seconds. During obstruction, character styling, held items, ground, and surrounding world must evolve simultaneously.
……