AI 视频生成器

Vampire Chase

模型

提示词

Prompt:
SCENE CONTEXT
A 15-second vertical 9:16 ultra-photorealistic live-action supernatural chase sequence. A young protagonist flees through a cold rain-soaked conifer forest while a vampire rapidly gains on them. The protagonist appears frightened and human until the final close-up reveals that they are also a vampire. Dark romantic thriller, realistic human anatomy, physically accurate full-speed running, cinematic but natural movement, fast continuous escalation, no prolonged slow motion.
ACTIVE REFERENCES — STRICT ROLE HIERARCHY
<<<image_1>>> — PROTAGONIST FACE: controls ONLY the protagonist’s exact facial identity, real apparent age, skin tone, and face details. Preserve the same recognizable person in every shot. Ignore the reference background, framing, lighting, and any non-face information. The eyes remain their natural reference color until the final transformation. Clothing is NOT controlled by the reference image and must be defined only by the prompt: if the input face is female, use a fitted black biker-style leather jacket, black top, black trousers, and black combat boots in a feminine cut and silhouette; if the input face is male, use the same outfit logic in a masculine cut and silhouette. Preserve the same hairstyle length and texture appropriate to the face input, rendered naturally in motion and rain.
<<<image_2>>> — PURSUER FACE: controls ONLY the pursuer’s exact facial identity, real apparent age, skin tone, and face details. Preserve the same recognizable person throughout the entire chase. Ignore the reference background, framing, lighting, and any non-face information. The pursuer’s eyes remain realistic dark burgundy-red throughout, with restrained vampire fangs. Clothing is NOT controlled by the reference image and must be defined only by the prompt: if the input face is female, use a black café-racer-style leather jacket, black top, black trousers, and black combat boots in a feminine cut and silhouette; if the input face is male, use the same outfit logic in a masculine cut and silhouette. Preserve the same hairstyle length and texture appropriate to the face input, rendered naturally in motion and rain.
LOCATION
Controls ONLY the rain-soaked conifer forest, split standing tree, wet running path, fallen trunk, narrow 1.5-meter rocky fissure, canyon overlook, gorge geography, wet materials, cold blue-gray dawn, rain, mist, and lighting. Preserve the same geographically continuous environment throughout the sequence.
REFERENCE SEPARATION
The protagonist and pursuer remain visually distinct at all times. Never exchange, merge, average, or blend their faces, hairstyles, bodies, clothing, or identities. Exactly one protagonist and one pursuer appear in the entire film. No additional people.
No studio backgrounds or reference-sheet layouts ever appear in the forest. The characters retain the exact clothing defined by the prompt.
FORMAT AND PACING
15 seconds, vertical 9:16, full realism. Seven fast cinematic beats connected by clear directional continuity. Each beat contains one primary action and one primary camera movement. Constant escalation. No calm resolution. Cut to black at peak tension.
0–1.5 SECONDS — THE HOOK
The protagonist stumbles urgently into frame and presses their back against the splintered side of the massive standing tree, breathing heavily after running. Rain beads on pale skin and the black leather jacket. Wet dark hair clings naturally to the cheeks. They suddenly raise their head after hearing movement behind them.
Camera: tight handheld side close-up with one slow push toward the face.
1.5–2.5 SECONDS — THE THREAT
Extreme macro close-up of one natural-colored eye snapping toward the sound. In the moist reflection of the iris, a dark figure moves rapidly between the distant tree trunks. The eye does not turn red yet.
Camera: one sharp hard cut into a static extreme macro close-up.
2.5–5 SECONDS — THE CHASE BEGINS
The protagonist explodes away from the tree and sprints at full speed along the wet forest path. The same vampire from <<<image_2>>> emerges from between the trees and immediately gives chase, remaining several steps behind. Their feet strike the muddy ground with realistic weight and throw small droplets backward.
Camera: fast low side-tracking shot racing parallel to both characters. Tree trunks and ferns create strong foreground parallax with controlled natural motion blur.
5–7 SECONDS — THE FALLEN TREE
The protagonist powerfully jumps over the fallen tree without slowing down, lands with believable weight, regains stride, and continues forward. The pursuer closes the distance and reaches one hand toward the protagonist’s back but narrowly misses.
Camera: one dynamic rear three-quarter tracking movement following closely behind them.
7–9 SECONDS — HE GAINS GROUND
The chase tears downhill between enormous wet trunks. The pursuer clears the same obstacle and rapidly reaches the protagonist’s shoulder line. The protagonist senses the pursuer and accelerates at the last possible moment. Both characters maintain correct running anatomy, ground contact, balance, and momentum.
Camera: one low handheld chase movement directly behind them, urgent but readable.
9–12 SECONDS — DIRECT PURSUIT
The protagonist runs straight toward the moving camera with fear and determination while the same pursuer remains clearly visible behind, rapidly gaining ground. The protagonist never poses or looks into the lens. The stride, arm movement, jacket movement, wet hair physics, and breathing remain anatomically natural.
Camera: one front-facing tracking shot moving backward at matching speed.
12–13.2 SECONDS — THE JUMP
The forest opens onto the wet rocky overlook beside the main canyon. The protagonist jumps across the narrow 1.5-meter fissure in the rocky ground near the canyon rim—not across the enormous main gorge. They clear the small gap and land hard in a controlled crouch on the other side. The pursuer stops at the near side of the fissure for a fraction of a second.
Camera: one forward sweep following the protagonist’s trajectory and dipping slightly with the landing.
13.2–15 SECONDS — THE REVEAL
Still crouched on the wet stone, the protagonist slowly raises their face toward the camera. Fear disappears. Both natural irises transform into realistic deep-crimson vampire irises with dark pupils and fine internal texture. A subtle confident smile forms, revealing two small upper canine fangs. The pursuer remains softly visible behind across the fissure, realizing that the protagonist is not helpless.
Camera: one rapid controlled push into an intense close-up. Cut to black at the peak of the expression.
CHARACTER PERFORMANCE
The protagonist begins genuinely frightened and exhausted: heavy breathing, tense shoulders, urgent glances, imperfect footing, and survival-driven movement. Fear gradually becomes focused determination. Only in the final moments does the expression become calm and predatory.
The pursuer is fast, controlled, and threatening. Does not snarl, roar, pose, or look into the camera. Remains focused exclusively on catching the protagonist.
PHYSICAL REALISM
Both characters maintain real ground contact, weight, balance, inertia, and forward momentum. Feet push against the terrain and never slide. Running is full speed but physically readable. Jumps follow natural takeoff, airborne trajectory, impact, and recovery. Clothing and hair react to rain, speed, gravity, and body movement.
No running in place, floating steps, weightless jumping, teleportation, sudden position swaps, obstacle phasing, or impossible speed changes.
GEOGRAPHIC CONTINUITY
The chase follows one continuous route through the location:
split standing tree → wet forest path → fallen tree → downhill path → rocky overlook → narrow 1.5-meter fissure beside the main canyon.
The narrow fissure is a small jumpable crack near the rim. It is never confused with the enormous canyon itself. Neither character attempts to jump across the full gorge.
VISUAL STYLE
Cold desaturated blue-gray palette, muted forest greens and natural browns, soft overcast dawn, fine rain, natural forest haze, wet skin and hair, realistic black leather, subtle film grain, crisp facial detail, shallow depth of field only in close-ups, documentary-grade handheld energy during the chase, controlled cinematic motion blur, high dynamic range, and readable shadow detail.
Ultra-photorealistic live-action cinema. No fantasy glow around either body, no game-rendered environment, no illustration, no cartoon appearance, no anime styling, no plastic skin, and no artificial CG sheen.
AUDIO
Heavy breathing, rapid wet footsteps, mud splashes, snapping branches, leather movement, rushing wind, steady rain, and distant forest ambience.
A deep restrained cinematic pulse begins when the pursuer appears and intensifies throughout the chase. Strong low-frequency impact during the final eye transformation. No dialogue, no voice-over, no captions, and no lyrics.
POSITIVE CONSTRAINTS
* Exactly one protagonist matching <<<image_1>>>.
* Exactly one pursuer matching <<<image_2>>>.
* One geographically continuous forest matching the location description above.
* Protagonist eyes remain natural until the final reveal.
* Pursuer eyes remain realistic dark burgundy-red throughout.
* Same faces, hairstyles, bodies, and clothing throughout.
* Full-speed physically believable running.
* Fallen-tree jump remains clearly visible.
* Final jump crosses only the narrow fissure, never the main canyon.
* Final crimson-eye and subtle-fang reveal remains clearly visible.
* Cut to black at the strongest final expression.
STRICT NEGATIVES
No face changes, identity drift, face swapping, blended identities, duplicate protagonist, duplicate pursuer, extra characters, costume changes, hairstyle changes, body distortion, altered height, deformed hands, extra fingers, duplicated limbs, floating movement, running in place, sliding feet, weightless jumps, teleportation, phasing through trees, jumping across the entire canyon, exaggerated superhero poses, snarling, monstrous mouths, oversized fangs, blood, gore, fantasy body glow, excessive slow motion, random camera cuts, studio backgrounds, character-sheet panels, text, captions, subtitles, logos, or watermarks.

10288 / 50000

参考图片（可选）

画面比例

时长

分辨率

生成音频2 倍积分Basic+

所需积分50

视频预览

加载中...