すべて · SNS投稿 · 共感ネタ

192言葉と本音

口にした返事と心の中の本音を、左右に並べた一枚でシェア。

項目を入力

1項目を入力します
2コピーを押します
3GPTのチャットに貼り付けます
4写真を使う場合は写真とプロンプトを一緒に送ります任意

画像ができたら：左右の文章と状況を確認し、一枚のまま投稿してください。

プロンプト全文を見る 3,185文字

Generate this image now. Draw it directly - do not describe it, explain it, or ask questions.
A3横、420×297mm、縦横比1.414:1のポスターを1枚作成してください。生成できる最大解像度を使い、印刷用の最終目標は4961×3508ピクセル、300dpiです。文字は紙の端から10mm以上内側に置いてください。

[Purpose]
Create one two-column Japanese social meme for one everyday situation, never a comic sequence or app screenshot.
[Scene]
Situation: "【状況】". Spoken reply: "【口にした言葉】". Inner thought: "【心の中】". If a reply equals "おまかせ", invent a short fictional Japanese line of at most twelve characters for that side. Otherwise print the exact user words. No invented real quotation or named-person attack.
[Attached photo]
No photo and no person. One clean speech bubble on the left, one pencil-scribbled thought bubble with two connected small circles on the right.
[Text layout]
Equal ice-blue left and muted-coral right halves divided vertically. Exact headings "言葉" and "本音" at y25-31%. Corresponding large replies at y40-59%,up to three short lines within each half. Exact situation once at y74-79%,centered without a box. Top22% and bottom18% completely empty. No extra title.
[Style]
Fixed matte editorial graphic with legible graphite pencil Japanese writing, off-white bubbles and slight paper grain. No portrait or gauge.
[Visual hierarchy]
Both replies equally dominant. Smaller headings and one quiet context line. No crossing the divider.
[Lighting and texture]
Flat paper and ink, no3D. Clean and scratchy contours give the contrast.

[Text rules]
Use only the requested visible copy in the target language. Resolve an automatic-copy instruction into an original short line; never print the instruction itself. Preserve user-supplied names, spelling, numbers and punctuation. Input labels, section headings and placeholders are not visible copy. Every optional field stays on its own instruction line so that leaving it empty removes only that item. Never add a date, price, discount, quantity, prize, origin, business condition, scientific statistic or claim about a real person that the user did not provide. Personal inputs affect content only; keep this one art medium, one layout and one finished result type.

[Avoid and delivery]
Return exactly ONE complete raster image directly. All requested elements belong on the same canvas. Do not run code, provide an explanation instead of an image, create separate panels as separate files, require another message, offer variants, or ask the user to cut and assemble anything. No device mockup, promotional wrapper, brand logo, packaging imitation, existing fictional character, artist signature, celebrity or watermark. Finish the selected canvas edge to edge. Maintain the fixed composition when the selected aspect ratio changes by adapting spacing, without cutting important parts or adding a second result.

[Legibility]
Keep essential lettering inside x7-93% and y14-84% of the image. The upper12% and lower15% contain no important text. At a1024x1536 cover, all essential ink is below y215 and above y1290. Use clearly connected natural letters, comfortable spacing and strong contrast. Do not let faces, decorative marks, food highlights or shadows obscure the words. Large lettering must still identify this result in a three-column list. Handwritten copy is legible pencil handwriting, never tiny ornamental cursive.