שלב 1: יצירת התמונה הראשונה
בוחרים דמות שאתם רוצים ומוצאים כמה תמונות רפרנס שלה (תמונות ברורות ובאיכות גבוהה, שה-AI ייעזר בהן). אחר כך מעתיקים את הפרומפט (ההוראה שכותבים ל-AI) שהכנתי לכם. בתוך הפרומפט מחליפים את [REFERENCE_SUBJECT] בשם של הדמות:
Vertical 9:16, hyper-realistic horror documentary photo.
A dark rectangular concrete drainage tunnel with damp walls, peeling paint, water stains, and patches of green moss and slime. The tunnel fades into deep black darkness behind the subject.
Center frame: [REFERENCE_SUBJECT] from the provided reference images, matched as accurately as possible (face/identity/clothing/pose fidelity), positioned as a half-body figure leaning slightly to the right and turned back toward the camera with a subtle unsettling expression. The subject appears dirty and weathered with realistic mud splatter and moisture stains, but still clearly recognizable as the reference identity.
Lighting: a single strong handheld flashlight beam from the left, creating a circular hotspot on the wall and across the subject, with harsh shadows, high contrast, heavy falloff into darkness, gritty low-light look, realistic sensor noise/film grain.
Camera: close-up, wide-angle perspective, handheld feel, sharp focus on the subject, background falling into black, no stylization, ultra-detailed concrete texture, realistic grime and moisture, cinematic found-footage realism, no CGI.
⸻
NEGATIVE PROMPT
cartoon, anime, illustration, 3d render, CGI, clean plastic look, glossy skin, beauty lighting, studio softbox, colorful neon, fantasy glow, perfect symmetry, smooth skin, low detail, blur, motion blur, overexposed, extra limbs, deformed face, wrong identity, inaccurate likeness, duplicated head, warped eyes, text, watermark, logo, frameעכשיו נכנסים ל-Nano Banana Pro (מחולל התמונות של Google). מעלים את תמונות הרפרנס, מדביקים את הפרומפט ומקבלים את התמונה של הטרנד:
שלב 2: הופכים את התמונה לסרטון
לא יצא בדיוק מה שרציתם? תשחקו קצת עם התוצאה עד שתקבלו את מה שאתם רוצים. אחר כך מעלים את התמונה ל-Veo 3.1 (כלי הסרטונים של Google). מדביקים את הפרומפט הזה, וזהו. יש לכם את הטרנד בסגנון שלכם.
POV video shot using a start frame and an end frame. The camera represents a real human point of view. The user is not just walking forward, but actively exploring and scanning the environment with the flashlight. The walking pace is very slow and cautious. The movement is grounded and realistic: Subtle vertical bob from slow, cautious footsteps. Gentle side-to-side sway. Imperfect handheld motion, showing slight hesitation. The user is holding a flashlight. The flashlight beam is the main actor in the scene, exploring the character: The beam doesn't just point forward; it actively scans. It moves up and down the character's body, from head to toe. The beam explores from different angles: The user slowly circles the character partially, or leans in and out, causing the light to hit the character's face and body from the side, below, and above, creating dramatic, shifting shadows. The movement of the light is natural and imperfect: Delayed movement, small overcorrections, uneven scanning, and slight shaking from the hand holding it. The light intensity on the character changes: As the user gets closer or the beam hits directly, the light is harsher (blown highlights); as it moves away or hits from an angle, the light is softer and shadows deepen. The character must remain completely still and lifeless at all times. No character animation. No blinking. No breathing. No movement. Only environmental motion is allowed: Floating dust motes illuminated by the flashlight beam. Minimal light flicker from the flashlight itself. Very subtle movement of loose elements (e.g., cobwebs, torn fabric) caused by air drafts. The environment must remain consistent throughout the entire shot. No set changes. No teleporting. No new objects. photorealistic video, found-footage realism, handheld walking and exploring camera, low light, strong dramatic shadows, gritty textures, natural imperfections, tense atmosphere.דוגמה לתוצאה שלי
הנה סרטון שיצרתי בשיטה הזאת: