Giữ mặt gốc, đổi tóc & makeup bằng 3 ảnh reference
Làm đẹp Hướng dẫn AI Styling · AChau AI
OUTPUT FORMAT — NON-NEGOTIABLE
Single continuous portrait frame. NO split-screen, NO grid, NO multi-panel, NO triptych. One unified image only.
FACE LOCK — PRIORITY OVERRIDE · Reference: @image_1
This is the ONLY accepted face and identity for this portrait. Match @image_1 precisely: face shape, eye shape and spacing, nose bridge and tip shape, lip shape and fullness, eyebrow shape, jawline, cheekbone placement, exact skin tone. Facial identity fully preserved — same person as @image_1, not a different face. Do NOT blend identity with @image_2 or @image_3 reference people — those are style-only references, their faces are discarded entirely.
HAIR — RESTYLED PER @image_2 · Reference: @image_2
Replace @image_1's original hair styling with the cut, length, texture and finish shown in @image_2: short tousled bob, wet-look texture with separated defined strands, side-swept fringe grazing the brow, slightly damp glossy sheen catching light, soft natural wave at the ends. Hair color stays @image_1's original dark near-black. Hair silhouette and styling follow @image_2 reference, not the original @image_1 hairstyle.
MAKEUP — LAYOUT PER @image_3 · Reference: @image_3 — technique and placement only
Discard that reference's own face/identity completely. Apply this makeup placement onto @image_1's face: Base — dewy, luminous, smooth and even, natural skin pores/texture preserved, not plastic-smoothed. Blush — soft warm pink-terracotta wash across cheeks, nose bridge and upper cheekbones, diffused soft edges, sun-kissed effect. Eyes — soft neutral-tone eyeshadow, subtle definition, natural lash flare, no heavy liner. Brows — soft groomed natural shape, not overdrawn. Lips — glossy nude-pink, high-shine finish, natural lip fullness. Overall: precise high-end fashion editorial makeup — clean, refined, not costume-heavy.
LIGHTING
High-end fashion editorial lighting. Large soft key light, front-diagonal angle, even wrap across the face. Clean, crisp micro-contrast on skin texture — no flat wash, no harsh hard shadow. Bright, clearly visible catchlights in both eyes. Color temperature: neutral-clean, 5000–5600K, true-to-life skin rendering. Background: soft neutral seamless, light beige/cream, minimal, gentle soft falloff at edges.
COMPOSITION / CAMERA
Portrait framing, chest-up to crown. Slight ¾ angle — avoid a flat frontal mugshot look. Eyes engaged naturally toward camera — alive, soft focus of attention, not a frozen AI stare, not theatrical. Shallow depth of field — face and eyes tack sharp, background gently soft. No AI over-smoothing on skin; texture stays visible under the makeup.
STYLE
High-end beauty/fashion campaign photography. Clean, crisp, commercial-grade finish — natural skin texture retained, not over-processed or waxy. Wardrobe: same neutral top visible in @image_1 reference, unchanged. No text, no logos, no watermark, no extra jewelry. Aspect ratio: 3:4 vertical portrait.
01 Chuẩn bị 3 ảnh reference đúng vai trò
Face: rõ mặt, ánh sáng đều, không filter, không che nét mặt. Hair: chỉ lấy form/độ dài/texture/mái, không lấy mặt người trong ảnh. Makeup: chỉ lấy base/blush/mắt/chân mày/môi, không lấy mặt người trong ảnh.
02 Set reference trong Higgsfield, model Seedream 4.5
Upload 3 hình, tag đúng @image_1 = Face/Identity, @image_2 = Hair, @image_3 = Makeup. Chỉ upload ảnh rồi prompt chung chung, AI dễ blend nhầm gương mặt của nhiều người.
03 Khoá nhận diện bằng đoạn Face Lock ưu tiên ghi đè
Mở đầu prompt bằng đoạn chỉ nhận @image_1 làm gương mặt — khớp chính xác hình dáng mặt, mắt, mũi, môi, chân mày, xương hàm, tông da; discard hoàn toàn gương mặt của @image_2 và @image_3.
04 Restyle tóc và layout makeup, viết tách hai đoạn riêng
Đoạn tóc mô tả cắt/độ dài/texture theo @image_2, giữ nguyên màu tóc gốc của @image_1. Đoạn makeup mô tả base/blush/mắt/mày/môi theo kỹ thuật của @image_3, không lấy gương mặt của ảnh đó.
05 Chốt ánh sáng, khung hình và ép output một ảnh liền
Soft key light chéo trước mặt, góc ¾, depth of field nông. Ghi rõ "one unified image, no split-screen/grid/triptych" ở đầu prompt để Seedream không tự chia khung.
ẢNH MINH HOẠ QUY TRÌNH 
FACE · @IMAGE_1 
HAIR · @IMAGE_2 
MAKEUP · @IMAGE_3 HiggsfieldSeedream 4.5
Ảnh · AI Styling Trial 
