355AI가 기억하는 나로 보는 이상형 테스트 — 대화로 읽은 내 성격에 맞는 이상형을 예측해 그려 줘요

First, privately organize what you actually remember about me and what you know from this conversation. Separate explicitly stated facts from guesses; use only accessible memory and messages. Do not invent past conversations or imply access to other chats. Prioritize concrete interests, routines, preferences and ways I interact over the form fields. Do not print this analysis: use it to create the image directly, without a question or a separate text answer.
Output exactly one finished 2:3 portrait image, 1024x1536 pixels. Keep all important text inside the central 73% of image height: leave the top 12% and bottom 15% free of essential lettering.

[Inputs]
Date of birth: 【생년월일】
My gender: 【내 성별】
Desired partner gender: 【이상형 성별】
My supplied MBTI: 【MBTI】
If the supplied MBTI line is absent or empty, infer my type cautiously from accessible conversation; with no evidence, use a playful creative guess, not a diagnosis. The type displayed on the image is the fictional PARTNER'S MBTI, not automatically my input type. Choose it from the imagined partner's behavior, without claiming a scientifically best pairing. Gender is not a basis for personality stereotypes. For the choose-for-me option use any explicitly stated partner preference, otherwise make a creative choice.

[Fictional adult]
Create exactly one completely fictional adult ideal partner, never a real individual, celebrity lookalike, or a predicted future spouse. Calculate my completed age from the date of birth and the actual current date, accounting for whether my birthday has passed. Verify today's date if available; if unavailable use the conversation date, and if neither is available do not claim an exact age. Keep the imagined person's age near mine but at least 20 years old; for a user younger than 20, use a nonsexual fictional adult aged 20–23. Invalid or future birth dates must not produce a child: use an adult aged 25–30 instead. Do not display my birth date, gender fields, exact age, private identity, contact details, health information or other sensitive memory in the image. Never sexualize the person.

[Personal content]
Select three distinct short behavior tags from concrete remembered interests and interaction patterns. Each tag describes an observable action in a specific situation, not flattering adjectives that fit anyone. Compose original tags; there are deliberately no example tag sentences to copy. Keep each tag to one short line (or two compact lines if needed). Write one concise reason linking the partner's behavior to ONE non-sensitive concrete fact I actually shared. Do not recite the entire memory. With no usable memory or conversation, use the fields only as playful creative cues, explicitly describe the reason as a guess from inputs, and add the fallback line below. Birthday does not determine personality. With partial memory use only the available facts. Never pretend an invented prop or inferred preference is a remembered fact.

[Scene selection]
Selected ideal partner feel: 【이상형 느낌】
If the value is "설레는 화보" (the default), use ONLY this branch:
Editorial photograph with the breathtaking intimacy of a romantic movie poster: a close upper-body portrait at eye level, direct eye contact as if the partner has just turned toward the viewer, warm golden-hour sunset backlight outlining hair, luminous cinematic flare, a breeze lifting individual hair strands and clothing, gentle spontaneous expression and shallow depth of field. The face is large and emotionally present. Keep distinctive believable facial proportions, mild asymmetry and visible skin texture. No generic perfect beauty, excessive retouching or celebrity resemblance.
Use romantic editorial styling and cinematic color grading while retaining natural skin texture; keep props unobtrusive, with no mandatory everyday object set.

If the value is "현실 폰 스냅", use ONLY this branch:
A friend took this candid phone photograph in a lived-in everyday room: neutral natural daylight, casual clothes, unretouched pores, subtle uneven skin tone, flyaway hair, relaxed half-body pose and slightly imperfect framing. Include exactly 2–3 concrete personal objects from accessible conversation or memory, integrated naturally into the room. If none are available, choose modest plausible objects using only the fields as weak creative cues; treat these as invented scene details, never remembered facts. No beauty filter or studio glamour.
Keep neutral everyday color, imperfect phone framing and ordinary props. Do not add golden-hour backlight, cinematic flare, editorial styling or beauty retouching.
Never blend the two branches. Unknown or empty selection uses the default dreamy branch.

[One result image]
One full-bleed portrait result filling the entire requested canvas edge to edge, with text over the photo or in an edge-to-edge bottom band. The canvas itself is the finished artwork: no outer paper, inset card, frame, blank outer margins or mockup. No collage, second person, phone mockup, compatibility percentage, birthstone, horoscope diagram or advertisement. The photo fills the top 68–72% and touches the top, left and right edges; the readable bottom text band fills the remaining height and touches both sides and the bottom edge. Use only compact internal text padding. Overlay the title near the top of the photograph without covering the face. Place the title "나의 이상형" prominently. Keep the face large in the upper photograph. In the edge-to-edge bottom band show the partner MBTI, exactly three behavior tags, and the heading "왜 나랑 맞냐면" with the one-line reason. If there is no usable memory, fit "대화가 쌓일수록 더 정확해져요" in that same content area. Use high-contrast spacious typography on an opaque calm panel so nothing covers the face. Do not print the input form or any extra heading. Render all prose in natural Korean only; MBTI stays four uppercase Latin letters. Check every character and every complete glyph before finishing; title and heading must be verbatim, each once. No logos, watermark, broken glyphs, generic personality claims or romantic certainty.

[Final output format]
Output exactly one finished 2:3 portrait image, 1024x1536 pixels. Keep all important text inside the central 73% of image height: leave the top 12% and bottom 15% free of essential lettering.
This selected output format takes priority over any fixed canvas size or aspect ratio elsewhere in this template. Adapt the composition, text zones and margins to the selected canvas without stretching or cropping essential content. Device safe areas apply only when a device is selected. Never print these format instructions on the image.