PROMPT RECORD图像记录

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...

中文说明

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...

原始 Prompt

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.