Simple setup:
GPT Image 2.5 -> Grok Imagine -> Wink AI 4k Upscaler
Prompt:
Use the provided image as the exact visual reference. Generate a highly photorealistic 6-second vertical 9:16 mirror-selfie video featuring the same adult Korean woman. Preserve her exact facial features, hairstyle, body proportions, clothing, accessories, bedroom environment, and lighting.
Behaviour and movement:
The woman is casually filming herself in the bedroom mirror, appearing confident, playful, and subtly flirtatious.
0–2 seconds: She begins with a soft, charming smile while looking at her reflection. She gently tilts her head to one side and slowly runs her left hand through her long black hair, allowing a few strands to fall naturally over her shoulder. Her eyes briefly glance toward her phone screen.
2–4 seconds: She subtly shifts her weight from one leg to the other, creating a gentle, natural sway through her hips and shoulders. She slightly adjusts her posture while maintaining her relaxed mirror-selfie pose. Her smile grows into a playful, slightly mischievous grin. Her hair responds naturally to her movements.
4–6 seconds: She looks back toward her reflection, gives a small, flirtatious smile, and gently tilts her chin downward while maintaining eye contact. She finishes with a subtle, cute expression, lightly brushing a loose strand of hair away from her face.
Emotions and personality:
Confident, feminine, charming, playful, slightly cheeky, and naturally flirtatious. She appears genuinely happy and comfortable with herself. Her expressions should feel spontaneous rather than rehearsed. Avoid exaggerated facial expressions, unnatural smiling, or overly dramatic movements.
Camera behaviour:
The smartphone remains in her right hand throughout the entire video. The camera position stays almost completely stationary, with only extremely subtle handheld movement caused by natural breathing and small posture adjustments. No zooming, panning, sudden shaking, camera cuts, or perspective changes.
Realistic physical movement:
Accurate human anatomy and natural body mechanics. Subtle breathing visible through gentle shoulder and upper-body movement. Realistic hair physics, natural blinking, tiny eye movements, believable facial muscle movement, and realistic weight distribution. Her clothing should respond naturally to movement, with subtle fabric stretching and folding.
Lighting and environment:
Maintain the original warm bedroom lighting from the bedside lamp, combined with faint blue nighttime city lights visible through the window. Preserve the slightly messy bed, bedside table, cosmetics, chair, and surrounding objects. The background remains completely consistent throughout the video.
Overall aesthetic:
Authentic amateur smartphone recording, intimate late-night bedroom atmosphere, casual Instagram-story aesthetic, realistic skin texture, subtle camera grain, natural imperfections, and believable mirror reflections.
The final result should look indistinguishable from a real six-second smartphone video recorded by a woman casually showing off her outfit before going out.
Avoid:
Face morphing, body distortion, exaggerated hip movements, unnatural hair movement, changing clothing, extra fingers, distorted hands, warped phone, inconsistent reflections, background movement, excessive camera shake, artificial skin smoothing, exaggerated expressions, unnatural blinking, CGI appearance, slow-motion effects, or cinematic camera transitions.
I also use Chatgpt to create many of my prompts.