请务必在为提示词生成关键帧时使用 GPT2 image，然后再将其推送到 Seedance 2.0 工具中。

Do well to always generate your keyframes with GPT2 image for your prompt before pushing it to Seedance 2.0 tools. This will let you spot out errors if there's any in the prompt and understand the feedback they give when plugged in. You can edit stuffs in the process. Also let you check and guide the consistency of your character. I made this one with this prompt 👇 A cinematic 2D sakuga anime action scene of the character standing in the middle of a futuristic urban street at bright daytime. He wears dark modern fantasy armor with subtle silver details and holds a glowing katana infused with blue energy. The camera starts with a dynamic dutch-angle medium-wide shot as the warrior suddenly launches forward through the air toward the camera at extreme speed. His red cape violently flutters behind him while blue energy sparks erupt from the katana. At the 2-second mark, the camera rotates slightly overhead as he swings the katana in a powerful arc slash. A bright crescent-shaped blue energy trail cuts across the screen with intense motion blur and fluid hand-drawn smear frames. The animation becomes highly fluid sakuga style with detailed impact frames, speed lines, animated debris, wind distortion, and dramatic perspective warping. The warrior’s expression is fierce and focused, mouth open as if shouting during the attack. The final shot ends with him landing on one knee as the cape settles slowly behind him while glowing energy particles drift through the air. Style: high-budget Japanese anime film, fluid sakuga animation, dynamic camera movement, motion blur, dramatic lighting, clean linework, highly detailed effects animation, modern anime movie quality, 24fps, cinematic composition. Constraints: - only one character visible - only one sword - no civilians - no extra limbs - no duplicated objects - maintain consistent face and outfit - maintain cape color consistency - smooth anatomy during motion - no camera shaking beyond cinematic movement - no 3D CGI appearance - no photorealism - preserve clean anime linework - use consistent daylight lighting - avoid blurry face during action

请务必在为提示词生成关键帧时使用 GPT2 image，然后再将其推送到 Seedance 2.0 工具中。 这样做可以让你发现提示词中是否存在错误，并理解插入后工具给出的反馈。 你可以在过程中进行编辑。 同时还可以检查并引导你角色的一致性。 我使用以下提示词制作了这个 👇 一个电影感的 2D 作画（sakuga）风格日本动漫动作场景，角色站在明亮的白天未来主义城市街道中央。他身穿带有微妙银色细节的深色现代奇幻盔甲，手持一把注入蓝色能量的发光武士刀。 镜头以一个动态的荷兰角（dutch-angle）中远景开始，战士突然以极高速度向前腾空扑向镜头。他的红色斗篷在身后剧烈翻飞，同时武士刀上迸发出蓝色能量火花。 在第 2 秒时，镜头略微旋转至俯视角度，他挥舞武士刀做出一次强力弧形斩击。一道明亮的蓝色新月形能量轨迹以强烈的运动模糊和流畅的手绘残影帧横切过整个画面。 动画变得高度流畅的作画（sakuga）风格，带有精细的冲击帧、速度线、动态碎片、风扭曲以及戏剧性的透视变形。战士的表情凶猛而专注，嘴巴张开仿佛在攻击中呐喊。 最后一个镜头以他单膝落地结束，斗篷在他身后缓缓落下，发光的能量粒子在空中飘散。 风格：高预算日本动漫电影，流畅的作画（sakuga）动画，动态的镜头运动，运动模糊，戏剧性的打光，干净的线条，极其精细的效果动画，现代动漫电影品质，24fps，电影感构图。 约束条件： - 仅可见一个角色 - 仅有一把剑 - 无平民 - 无多余肢体 - 无重复对象 - 保持一致的面部与服装 - 保持斗篷颜色一致性 - 动作过程中解剖结构平滑 - 除电影感运动外无镜头抖动 - 无 3D CGI 外观 - 无照片级真实感 - 保留干净的动漫线条 - 使用一致的日光照明 - 动作过程中避免面部模糊