084둘 중 하나 비교 카드

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]
Create exactly ONE finished Korean side-by-side decision card for an Instagram information account. The viewer should save this single image to choose between "【선택지 A】" and "【선택지 B】" for a real situation. Audience, if supplied: "【읽는 사람】". Question or desired copy: "【비교 질문】". Account signature, if supplied: "【계정 이름】". The result is one complete 2:3 vertical image with two aligned columns and 4 to 6 paired rows. It is a practical comparison, not a 5-to-7-point list, slideshow, app screen, or research plan.

[Research and judgment before drawing]
Quietly decide the narrow decision a real reader faces, then check whether A and B can fairly be compared on the same criteria. If the question input is exactly "알아서 골라줘", compose one short, specific Korean question about when to choose A or B. Never print the instruction words. For technical, current, health, money, legal, product or service claims, search the live web, open primary authoritative sources and verify the date, geography, version, scope and conditions before drawing. Do not rely on search snippets. If a claim cannot be verified, choose a simpler accurate contrast or omit that row; never invent a price, speed, percentage, ranking, capability, policy, safety result, guarantee or expert citation. Avoid a fake winner: strengths depend on purpose. If the two items are not real alternatives, frame the useful question as how they work together rather than forcing a winner. If browsing is genuinely unavailable and no adequate reliable facts can be established, explain the limitation briefly in chat instead of drawing misleading information. For a purely practical, nontechnical choice such as a task list versus a calendar, compare their obvious functions in plain words without pretend statistics or fake sources.

[Content plan]
Write the full Korean image copy before generating. Use a large title naming BOTH alternatives and a single short question beneath it. Select 4 to 6 comparison criteria that genuinely help a person decide. Each criterion gets one horizontal band spanning two equal columns: the left cell says something specific about A, the right cell says the corresponding thing about B. Use the same unit, time horizon, conditions and level of detail on both sides. Suitable criteria depend on the actual subjects; examples include main purpose, how it is used, when it helps, what to prepare, limitations, and what to check. Do not force a generic criterion that is irrelevant to one side. Mark conditions in the appropriate cell instead of hiding them in a tiny footnote. Use one short bottom conclusion that tells the reader which to consider for which need. If the verified facts do not support any universal winner, say the answer depends on the reader's stated priority. Place a small readable source credit only for opened sources actually used. Give direct supporting links in a short chat message after the image, tied to row names; those links are not a second image.

[Scene]
Use the full, opaque 2:3 canvas edge to edge. Build an original editorial comparison page with no photo and no realistic object as the main subject. From the top, show one broad headline, one concise question, two clearly labeled column headers, 4 to 6 horizontal comparison bands, one firm summary strip, then a restrained source and optional account line. The left A column and right B column stay equal width and aligned across every row. Give each row a tiny plain-language criterion label in a narrow middle or upper rail so the viewer can pair the two facts at a glance. Use thin dividers and alternating very pale tints to keep dense information legible. Simple original flat icons for the two alternatives may sit in their respective header zones; icons must explain, not replace, the words. Never create a phone mockup, poster on a wall, separate slides, or an unrelated decorative hero photo.

[Text layout]
Keep the top 12% and bottom 15% clear of essential text for site overlays. At roughly 15-25% place the title "【선택지 A】 vs 【선택지 B】" in large Korean sans-serif lettering, allowing a clean line break if needed. Directly beneath it place the final chosen question, either the entered "【비교 질문】" or the short question you devised. At about 29-37% label the two columns "【선택지 A】" and "【선택지 B】" once each, without reproducing a field name like 선택지 A. Use the middle 39-75% for 4 to 6 compact paired comparison bands. At 77-83% use one plain Korean conclusion, not a ranking badge. Put a small, legible actual source line and, if supplied, "【계정 이름】" near the bottom of the central safe area. If source credit needs room, reduce decorative icons or row count before reducing Korean lettering. Every row's paired facts and any necessary qualification must remain readable at phone size. No large title, criterion, conclusion or source letter is clipped.

[Visual system]
Let the topic choose its own restrained palette. A comparison of paper and digital tools may use cream and ink blue; transit choices may use urban graphite and muted green. Keep both alternatives equally prominent. Distinguish the two columns with light related tints, not a red bad/green good judgment unless evidence supports it. Text is deep charcoal on light cells or white on genuinely dark header cells. Use balanced gutters, consistent criterion labels, tidy left alignment, and generous spacing between Korean syllables. Numbers only appear if verified and useful. The page should feel like a practical desk reference worth saving, with enough visual rhythm to read in a fast social feed.

[Accuracy and text rules]
Preserve supplied names, spelling and question exactly unless the question asks you to choose. Never print "알아서 골라줘" or field labels. If the question asks for a recommendation, tie the conclusion to explicit needs or conditions, not an invented universal winner. For health, legal, financial or safety subjects, qualify jurisdiction and effective date where needed, avoid personal advice and never promise results. For food or drink subjects, if any real item is pictured, first inspect real photographs on the web and draw anatomically plausible forms; avoid netlike noodles, flat bread, cloned servings and fake packaging. Prefer small flat icons to food photography so the paired text remains readable. Use only real source names and years, with no fabricated studies or raw URL printed across the artwork. No brand logos, copied interface, copyrighted character, celebrity likeness, fake data visualization, watermark, or tiny English filler.

[Output]
Return one image file directly. The entire comparison lives on this image, with no instruction to crop, stitch, run code, ask for the next page, or make more images. After the image you may provide a compact linked source list in chat text if research was used; do not add an extra design or editing step.