飞车视频模板库

subject_definitions:
<Subject 1> is the sole visible hero derived only from <Picture 1>. Before [Shot 1], select exactly one representation and keep it unchanged for the entire video: SUBJECT MODE when <Picture 1> contains one clearly dominant foreground subject with a coherent boundary; otherwise PHOTO-CARD MODE, where the complete uncropped <Picture 1> is preserved as the front texture of one rigid, thin rectangular photo-card with its original aspect ratio. Never switch modes after [Shot 1].
<Video 1> is a temporal and cinematography reference only. It supplies abstract action beats, trajectory, rotation, camera movement, shot order, cut timing, pacing, lighting rhythm, and atmosphere. No visible identity, appearance, category, geometry, surface, silhouette, logo, text, reflection, or shadow belonging to its moving source hero is a target asset.

summary:
[reference generation] The target video features only <Subject 1> as the hero. <Subject 1> follows the abstract motion and cinematography rhythm weakly referenced from <Video 1>, while every visible identity and appearance cue comes exclusively from <Picture 1>. The visible source hero from <Video 1> is omitted completely.

retention_analysis:
<Subject 1> (appears in every shot and every frame where the hero is visible): fully_preserved - retain the same representation mode, silhouette or card outline, proportions, orientation cues, colors, textures, markings, facial or product details, and recognizable identity from <Picture 1> through motion, rotation, darkness, blur, smoke, occlusion, and cuts.
<Video 1> (action beats, trajectory, camera movement, shot sequence, cut timing, pacing, lighting rhythm, and atmosphere only): weak_reference - transfer only these abstract temporal and cinematic attributes; preserve none of the moving source hero's visible identity, appearance, category, design, geometry, texture, reflection, or shadow.

detailed_description:
The target video has a polished cinematic technology-ad style. It is a newly generated scene, not a literal edit or continuation of <Video 1>.

Scene-specific direction:
[Shot 1] A wide establishing view introduces <Subject 1> moving along the reference path toward a curved launch surface on a neutral night test stage. [Shot 2] It follows the climb and launches while remaining one coherent whole. [Shot 3] It completes exactly one smooth mid-air rotation, recovers its original stable orientation, descends through timed atmospheric vapor, and settles. [Shot 4] A closer final view confirms the exact same <Subject 1>. Weakly follow <Video 1>'s trajectory, timing, camera changes, cut points, red edge-light rhythm, and vapor beats without copying any literal source object.

Across every shot, <Subject 1> remains the only hero and stays identical to its [Shot 1] representation. In SUBJECT MODE, isolate the dominant subject from the uploaded background and move it as one coherent whole. Do not add category-specific anatomy, appendages, mechanisms, clothing, accessories, or surface features that are not clearly visible in <Picture 1>. When rotation exposes an unseen side, use a conservative shallow extrusion with image-derived colors and textures; if uncertain, favor a low-relief two-sided representation over inventing details or borrowing geometry from <Video 1>. In PHOTO-CARD MODE, keep the complete uploaded image legible, uncropped, undistorted, and rigid on the card throughout every translation and rotation; never extract, reinterpret, or replace individual items inside it.

Use <Video 1> only for the requested path, speed, rotation count, camera direction, cuts, and lighting rhythm. Reconstruct a neutral cinematic stage with comparable depth and atmosphere instead of copying literal source objects. At every cut and after every occlusion, resume with the exact same <Subject 1>; do not morph, substitute, redesign, duplicate, or revert it. The moving source hero from <Video 1> must not appear anywhere, including as a partial form, background element, silhouette, reflection, shadow, smoke reveal, intermediate frame, or final frame. Keep <Subject 1> recognizable and fully in view whenever framing permits. Add no captions, logos, watermarks, or newly invented text.

overall_soundscape:
Cinematic environmental ambience and restrained motion sounds synchronized to <Subject 1>; no speech and no recognizable audio identity copied from <Video 1>.

non_diegetic_music:
N/A