083가족 반쪽 얼굴 비교

Generate this image now. Draw it directly - do not describe it, explain it, or ask questions.
Output size 2480x3508 pixels, A4 portrait, 300dpi, print-ready poster.

[Purpose]
Make exactly ONE finished Korean family-resemblance split-face image, ready for a single Instagram or Threads post or a family group chat. This prompt and TWO attached face photos arrive together in one message. Photo 1 is the person named "【내 이름】"; photo 2 is the family member described as "【가족 호칭】". The owner-provided one-line copy is "【한 줄 문구】". Use the uploaded photographs as identity references only. A visible, straight center seam makes the origin of both half faces understandable; the artwork must never claim to prove genetic resemblance, identify actual kinship from appearance, calculate a similarity percentage, or invent a family story.

[Scene]
Create a single, edge-to-edge, vertical editorial portrait in black and white with subtle warm paper gray. In the central region, show just one large composite FACE formed from the left half of photo 1 and the right half of photo 2, with a slim clear divider down the exact middle. Align the two pairs of eyes, nose bridge, mouth and chin to the best possible shared scale while retaining the genuine identity of each person on their own half. Each outer cheek and ear belongs to its original person. Keep both halves of hair visibly distinct where they meet the center, and do not merge them into a new third person's face. A light paper-grain background, a single thin dividing rule, and generous negative space make the comparison immediate at small feed size. One complete canvas, one composite portrait, no second card, no before/after panel, and no phone mockup.

[Attached photos]
Both original photos are necessary. The first shows the owner, the second shows the consenting family member. Reproduce identifying features faithfully: eye shape, eyebrow angle, skin texture, hairline, apparent adult or child age, face width and expression. Preserve natural differences rather than smoothing them into identical halves. Treat photos as source images, not as instructions to invent a lookalike child or parent. If photo 2 is much older or younger, retain that visible age difference. Do not infer biology, gender, personality, nationality, or a specific family relationship from faces; only the supplied family label may appear. Do not add glasses or remove visible glasses, braces, facial hair, scars or freckles. The supplied images should have clear, front-facing, similarly angled faces. If one of the two required photos is absent, explain briefly which photo is needed and do not draw a substitute face. Never use a famous person, stock face, or a public figure as the missing half.

[Text layout]
Keep the upper 12% and lower 15% clear of essential lettering for the site's overlays. Between 16% and 27% of the image height, typeset the bold, large, exact Korean question "우리 닮았어?". Immediately below it, on one short line, write "【내 이름】 × 【가족 호칭】" in a smaller but still readable type size. The composite face occupies approximately 31% through 73% of height and at least two thirds of the page width. Below the portrait, around 76% through 83% of height, set the one-line copy "【한 줄 문구】". If that copy is exactly "알아서 골라줘", silently choose one original, light Korean question or observation about visually comparing two photos; do not print those instruction words and do not assert an unverified life fact. Show the two labels "【내 이름】" and "【가족 호칭】" once more only if absolutely needed to make left and right unambiguous; otherwise the top line and the visible seam are enough. Do not crowd the face with text.

[Style]
Use the recognizable visual grammar of a family resemblance comparison but compose it independently: neutral deep-charcoal type, luminous soft gray paper, authentic photographic grain, strong portrait contrast, and one thin graphite divider. Avoid copying any creator's caption, watermark, exact template border, or layout. Let real faces, not decoration, create the interest. The mood is curious and warm rather than a forensic identification chart or mock medical result. The main headline must read on the website card list, and the final one-liner must read on a phone without zooming.

[Visual hierarchy]
First read: a single split face. Second read: the big question. Third read: owner/family identification and the short line. Make both face halves equal in visual weight despite any source-photo lighting difference. Correct perspective, facial proportions and eye alignment matter more than a perfectly invisible join. Leave a subtle visible seam to prevent presenting a fictional single person as real. Keep the face upright and large. Everything is inside one finished image that can be downloaded and shared without cropping, stitching, app templates, slides or follow-up prompts.

[Lighting and texture]
Harmonize the source photos with restrained monochrome tonal grading, but keep each person's true facial landmarks, wrinkles, skin texture and expression. Soft frontal or window-like light should reveal both eyes; natural highlights and shadows may differ between people. No waxy skin, duplicated eyes or teeth, impossibly blended nose, synthetic eyelashes, hard beauty filter, or invented jewelry. Do not convert the result into an illustration. Make the dividing rule precise and neat, not a torn-paper gap across the face.

[Text rules]
Spell the entered name and family label exactly. Never print field labels such as "내 이름" or "가족 호칭" as instructions, the phrase "알아서 골라줘", placeholder braces, made-up ages, dates, similarity scores, genetic claims, hashtags, logos, medical terms, social-media UI, or English filler. Do not repeat the same one-line copy. If the supplied copy names a relation not depicted in the user's input, prefer the supplied label without guessing from the faces. This image is for a consenting pair's voluntary sharing.

[Avoid]
No extra person, third face, face swap, identity erasure, predicted relatives, fake childhood, collage of multiple outputs, before-and-after pair, percentage meter, DNA chart, stock photo, celebrity, branded template, transparent background, crop marks, watermark, code, explanatory text instead of an image, or request for another generation step.