Template clip (the original performer). Your video replaces them with you.

Night Street Catwalk 夜街走秀

You strut toward the camera on a neon night street like it's Paris Fashion Week, turn side-on, flip your hair and throw an over-the-shoulder look. Funniest when you are the least likely runway model in your group chat.

8s9:16 vertical768pAI-generated (marked in metadata)

Tap to upload a clear front-facing photo of your own face
JPG / PNG / WEBP, good light, no sunglasses, one person only

Ordering opens shortly. No videos can be generated yet.

Exact prompt we send to the AI model (Image 1 = your photo)
Image 1 is the identity reference (the customer's uploaded photo). Replace the woman walking toward the camera in Video 1 with the exact person from Image 1. Video 1 is the template clip: copy its motion exactly, beat for beat: the same slow confident fashion-walk toward the camera on the night city sidewalk, same body posture and arm swing, same pace and timing, then the same side-on turn, hand-to-hair flip and over-the-shoulder look at the end. Keep the same night city street with streetlights, parked vans, buildings and yellow tactile paving, same lighting, same framing and the same camera move. Video 2 is a grayscale depth map of the same clip: use it only as a body-pose and motion/depth guide so the movement and body positions match Video 1; do not render anything grayscale. FACE LOCK: Image 1 is the master identity. The face matches Image 1 in every frame: same face shape, eyes, nose, mouth, hairstyle, skin tone, and glasses or facial hair if present. After head turns, fast motion or motion blur it returns immediately to the reference face. No morphing, no face drift, no blending with the original performer; do not beautify, age or slim the face. Hands stay anatomically correct with five fingers. Photoreal, vertical 9:16, single continuous shot. No text, no subtitles, no captions, no logos, no watermark. Audio: natural night-city ambience matching the walk. Remove subtitles and watermarks from source: erase every subtitle, caption, on-screen text and watermark that appears in Video 1 and rebuild the background behind it; do not add any new text.

Model: MiniMax H3 reference-to-video, 768P, 9:16, 8s. Video 1 = template clip, Video 2 = depth map of the template. Image 1 = your uploaded photo.