3. 提示词的额外技巧

3. Extra TIPs for prompting Do a rough prompt first with your own words. Then ask an LLM to rewrite the full prompt so Seedance has a better clarity of it. I find Chatgpt works better for this task than Gemini. You can ask spcifically to the LLM to write a prompt to describe the type of animation you want. You can even upload the video to Gemini/ChatGpt and ask it to analyze it and then extract the componets you are interested in (specific animatinos, colours, framing, camera movement, even the overall video) Example: Input for ChatGPT: "Write a prompt to describe an animation of Hunter x Hunter Hisoka, where he moves at high speed and appeares as an elongated mas of smearead blurry body. It appears in varoous spots in the same shot. Adapt the text prompt for Seedance 2.0 understanding" Output: "Create a dynamic anime action shot of Hisoka from Hunter x Hunter moving at extreme speed across the frame. His movement is so fast that his body stretches into a long distorted smear, like an elongated blurred mass of color and motion, with strong directional deformation. He appears in multiple positions within the same shot due to his speed, leaving behind successive afterimages and motion trails that clearly show his path through space. The effect should feel like high-end hand-drawn anime speed animation: stylized smear frames, stretched anatomy, motion blur, duplicated positions in one frame, rapid displacement, and aggressive visual momentum. Keep Hisoka recognizable in silhouette, costume colors, and overall design, but emphasize speed distortion over clean body detail during the fast movement. Style: premium modern anime, sharp cinematic composition, dramatic action energy, stylized 2D smear animation, strong motion readability, high-speed teleport-like movement, dynamic afterimages, impact-driven staging. Important: - show multiple visible positions of the character in the same shot - body becomes a stretched blurry smear during the fastest motion - preserve clear direction of travel - use anime-style deformation, not realistic live-action blur - make the movement feel violent, instant, and superhuman - keep the shot readable and visually striking"

3. 提示词的额外技巧 先用自己的话写一个粗略的提示词。然后让一个大语言模型（LLM）把完整的提示词改写一遍，让 Seedance 能更清楚地理解。 我发现 ChatGPT 在这个任务上比 Gemini 效果更好。 你可以专门让 LLM 写一段提示词，来描述你想要的那种动画类型。你甚至可以把视频上传到 Gemini/ChatGPT，让它分析视频，然后提取你感兴趣的各个组成部分（具体的动画、配色、构图、镜头运动，甚至整个视频的总体风格）。 示例： 给 ChatGPT 的输入："Write a prompt to describe an animation of Hunter x Hunter Hisoka, where he moves at high speed and appeares as an elongated mas of smearead blurry body. It appears in varoous spots in the same shot. Adapt the text prompt for Seedance 2.0 understanding" 输出："Create a dynamic anime action shot of Hisoka from Hunter x Hunter moving at extreme speed across the frame. His movement is so fast that his body stretches into a long distorted smear, like an elongated blurred mass of color and motion, with strong directional deformation. He appears in multiple positions within the same shot due to his speed, leaving behind successive afterimages and motion trails that clearly show his path through space. The effect should feel like high-end hand-drawn anime speed animation: stylized smear frames, stretched anatomy, motion blur, duplicated positions in one frame, rapid displacement, and aggressive visual momentum. Keep Hisoka recognizable in silhouette, costume colors, and overall design, but emphasize speed distortion over clean body detail during the fast movement. Style: premium modern anime, sharp cinematic composition, dramatic action energy, stylized 2D smear animation, strong motion readability, high-speed teleport-like movement, dynamic afterimages, impact-driven staging. Important: - show multiple visible positions of the character in the same shot - body becomes a stretched blurry smear during the fastest motion - preserve clear direction of travel - use anime-style deformation, not realistic live-action blur - make the movement feel violent, instant, and superhuman - keep the shot readable and visually striking"