Before After Morph

Before-After Morph

Produce smooth morph videos that sell the emotional impact of a simulated outcome — ideal for iPad 咨询室展示.

Pipeline

1. Align both images via eye-nose anchor similarity transform
2. Cross-dissolve 0–100% over 2–3 seconds (simple)
3. Mesh morph landmark-interpolated warp (premium)
4. Export MP4 + GIF + WebM

Alignment (required)

Misaligned morphs look like lighting changes, not structural change.

import cv2
import numpy as np

def align_faces(before, after, src_pts, dst_pts):
    M, _ = cv2.estimateAffinePartial2D(src_pts, dst_pts)
    h, w = before.shape[:2]
    return cv2.warpAffine(after, M, (w, h))

Anchor points: left eye center, right eye center, nose tip (stable across warps).

Morph Types

Export Specs

• Resolution: match before image, max 1080p for display
• Duration: 2.5s loop (before→after→hold→reverse optional)
• FPS: 30
• Add corner label: SIMULATION

Script

python scripts/morph_sequence.py before.jpg after.jpg --output morph.mp4 --duration 2.5
python scripts/morph_sequence.py before.jpg after.jpg --output morph.gif --fps 15

Creative Variations

• Split-screen wipe: after image wipes left over before
• Slider export: generate 10 frames for interactive before/after slider UI
• Slow reveal: mask expands from nose region outward (rhinoplasty reveal)

See examples.md for ffmpeg post-processing (fade, logo overlay).