How it works
One-Tap Makeup does not regenerate, reshape, or smooth the portrait. It requires one complete near-front face, cross-checks dedicated face detection with MediaPipe Face Mesh V2, and uses MediaPipe six-class portrait segmentation to distinguish face skin from hair, beard, body, clothing, accessories, and background. The preset creates exactly five compact cosmetic zones: two above the upper eyelids, two over the upper cheeks, and one inside the outer lip contour. Eye, iris, eyebrow, and inner-mouth pixels are explicitly protected. Blush and lip tint change LAB chroma while retaining source luminance; eyeshadow adds chroma with no more than four LAB luminance levels of darkening at full intensity. Facial geometry, pores, fine skin texture, expression, hair, beard, clothing, lighting, background, framing, uploaded alpha, and every pixel outside the reported zones are copied unchanged into a lossless PNG. Profiles, incomplete faces, uncertain face-skin evidence, faces too small for safe contours, and any glasses or obstruction that leaves a zone without sufficient safe skin support are refused instead of redrawn.
