我已附加一段参考视频和一份角色设定表风格参考。请按三步流水线执行。不要跳过任何步骤。任何输出中都不要添加文字或字幕。
分享链接 · https://openimages.ajiang.me/p/2081745761738412406/

color_palette · mutedcomposition · full-bodycomposition · multi-panelhuman_attributes · long-hairlighting · studio-lightsubject_type · character
提示词(原文,逐字保留)
I have attached a reference video and a character sheet style reference. Execute this as a three-step pipeline. Do not skip steps. Do not add text or captions anywhere in any output.
Step 1: Character Generation
Study the character sheet style reference I have attached. Replicate that exact character sheet format (pose layout, framing conventions, neutral background, consistent lighting, multi-angle presentation) to design the rider seen in the attached reference video. Match the rider's build, wardrobe, gear, and demeanor as observed in the video. Output a full character sheet in that style so the rider can be re-used with identity consistency downstream.
Step 2: Initial Image Generation (Seedream 5.0 Pro, 2K)
Using the character from Step 1 as the identity lock, generate the opening frame of the video I attached. Analyze the reference video's composition, environment, lighting direction, subject scale, and camera angle, and replicate that opening beat precisely.
Render two versions of this initial image: one at 9:16 and one at 16:9. Reframe the composition intelligently for each aspect ratio rather than cropping.
Visual specification for both images:
Extreme photorealism. High-speed tracking/panning photography aesthetic, frozen at the opening instant. Extreme camera angle matching the reference video's opening shot. Correct biomechanics of rider and machine. Detailed features, refractions and reflections on gear, metal, glass, and any wet or polished surfaces. Strong parallax between foreground rider and background environment. Luminous mist drifting far beneath the subject. Warm sunrise or sunset light as the key, with rich saturation across the frame. Fine 35mm film grain, vintage color grade, gentle bloom on highlights, subtle halation around bright light sources, prismatic highlights, controlled chromatic aberration at frame edges.
Step 3: Video Generation (Seedance 2.0, 1080p, 10 seconds)
Generate two videos, one at 9:16 and one at 16:9. For each, use the corresponding aspect-ratio image from Step 2 as the starting frame, and use the attached reference video as the motion guideline input so the animation, camera choreography, timing, and dynamics match the reference beat-for-beat across the full 10 seconds.
Motion and camera direction:
Replicate the reference video's motion in detail. Extreme camera angles. Maximum speed and dynamics. High-speed tracking and panning that follows the rider through the scene. Pronounced motion blur on fast-moving elements. Strong parallax between rider, mid-ground, and far background. Biomechanically correct rider movement throughout.
Look and grade, held consistently across all 10 seconds:
Extreme photorealism. Warm sunrise or sunset key light with rich saturation. Luminous mist far beneath the subject, drifting with parallax. Refractions and reflections that evolve naturally with camera motion. Fine 35mm film grain, vintage grade, gentle bloom, subtle halation, prismatic highlights, controlled chromatic aberration.
Deliver: character sheet (Step 1), two 2K images at 9:16 and 16:9 (Step 2), two 1080p 10-second videos at 9:16 and 16:9 (Step 3). No text, watermarks, or captions in any output.
中文译文
我已附加一段参考视频和一份角色设定表风格参考。请按三步流水线执行。不要跳过任何步骤。任何输出中都不要添加文字或字幕。
第一步:角色生成
研究我附加的角色设定表风格参考。完整复刻该角色设定表的格式(姿势布局、取景惯例、中性背景、一致的灯光、多角度呈现),用以设计参考视频中出现的骑行者。匹配视频中所观察到的骑行者体型、衣着、装备与气质。以该风格输出一份完整的角色设定表,以便骑行者可在后续流程中以身份一致性的方式重复使用。
第二步:初始图像生成(Seedream 5.0 Pro,2K)
以第一步中的角色作为身份锁定,生成我附加视频的起始帧。分析参考视频的构图、环境、光线方向、主体尺度与镜头角度,并精确复刻该起始节拍。
将这张初始图像渲染为两个版本:一张 9:16,一张 16:9。请针对各画幅智能重新构图,而非简单裁切。
两张图像的视觉规格:
极致照片级真实感。高速跟拍/摇拍摄影美学,定格于起始瞬间。匹配参考视频起始镜头所采用的极致镜头角度。骑行者与机械之间的生物力学正确。装备、金属、玻璃以及任何湿润或抛光表面的细节特征、折射与反射。骑行者前景与环境背景之间强烈的视差。主体下方远处飘动的发光薄雾。以暖色日出或夕光作为主光,全帧饱和度浓郁。细腻的 35mm 胶片颗粒,复古调色,高光处的柔和辉光,明亮光源周围的轻微光晕,棱镜高光,画幅边缘受控的色差。
第三步:视频生成(Seedance 2.0,1080p,10 秒)
生成两段视频,分别为 9:16 与 16:9。每一段以第二步中对应画幅的图像作为起始帧,并以附加的参考视频作为运动指引输入,使动画、镜头编排、节奏与动态在完整的 10 秒内逐拍匹配参考。
运动与镜头方向:
详尽复刻参考视频的运动。极致镜头角度。最大化的速度与动态。高速跟拍与摇拍贯穿场景跟随骑行者。快速运动元素呈现明显的运动模糊。骑行者、中景与远景背景之间强烈的视差。贯穿全程的骑行者生物力学正确动作。
全程一致保持的画面与调色:
极致照片级真实感。暖色日出或夕光主光,饱和度浓郁。主体下方远处的发光薄雾,随视差飘动。折射与反射随镜头运动自然演变。细腻的 35mm 胶片颗粒,复古调色,柔和辉光,轻微光晕,棱镜高光,受控的色差。
交付内容:角色设定表(第一步),两张 2K 图像(9:16 与 16:9)(第二步),两段 1080p、10 秒视频(9:16 与 16:9)(第三步)。任何输出中均不得出现文字、水印或字幕。
来源与署名
原文由 @maxescu 发布在 X:查看原推文。 本页逐字保留原文并提供机器翻译的中文解读;版权归原作者所有。
同工具的更多提示词
- [SUBJECT],超写实 3D 渲染,光滑充气光泽塑料材质,柔软蓬松气球般的轮廓,若隐若现的车缝拼接线和分块,高光镜面反射,干净的反光,柔和的接触阴影,设计…@Naiknelofar788
- 超写实 RAW 美女肖像,以上传的图片作为精确的面部参考。东南亚年轻女性,二十出头,柔和的鹅蛋脸,白皙象牙色暖中性肌肤,带有真实的毛孔和自然的水光肌光泽,浓密…@Naiknelofar788
- 一张高度写实、亲密的夜间卧室自拍,照片中一位年轻成年东亚女性舒适地躺在床上,以一种随意、自发的智能手机拍摄方式捕捉,拍摄于即将入睡的时刻。请使用上传的图片作为…@Naiknelofar788
- 目的地]: 一张高度细致的照片级真实感、值得分享到 Instagram 的旅行明信片立体模型,主体为 [DESTINATION],采用略微抬高的视角观赏。前景…@Naiknelofar788
- 为{DESTINATION}创作一张优雅的3D衍纸工艺(paper quilling)与多层纸质手工艺旅行海报,呈现高端复古旅行剪贴簿风格。艺术化地再现该目的…@Naiknelofar788
- 目的地]: 一张高度精细、照片般逼真的 Instagram 级别旅行明信片立体模型,主题为[DESTINATION],从略微偏高的视角观看。画面前景是一张真实…@aatif_j