
原文
GPT Image 2 · en
Scene:
A square survey-oriented image that explains spatial logic with the clarity of an expert field overview while still reading as a finished artwork. Use a stable elevated vantage, clean depth layering, and enough visible terrain or architecture to make relationships legible without turning the frame into a literal map. The scene should feel analytical and cinematic at the same time.
Subject:
Show a measured room perspective that explains wall programs, floor pattern, doorway placement, and the relationship between fresco surfaces and architecture. Organize the composition so paths, terraces, walls, courtyards, chambers, canals, root systems, or quarry clusters can be read as a coherent system. The image should reveal how the civilization used land, water, elevation, enclosure, or monumentality. Keep all architecture, materials, and environmental details consistent with the correct place and period.
Important details:
Keep the room recognizably Minoan, with fluid marine motifs, elegant fresco cadence, red and blue pigment behavior, and Bronze Age architectural forms rather than later classical Greek systems. This image should reward close looking by showing infrastructural logic: how stone courses align, how water moves, how rooms connect, how carved masses sit inside geology, or how vegetation interacts with built form. Avoid explicit annotation. Let composition, light, and scale cues do the explanatory work. Keep any measuring elements generic and unreadable, and make the image feel like serious research made visually elegant.
Use case:
Minoan Frescoes Living Again single study 8/9.
Constraints:
No readable text, no watermark, no logos, preserve exact 1:1 aspect ratio, no Roman statues, no Renaissance furniture, no fantasy palaces, no modern gallery labels, no infographic arrows, no split-screen layout, no impossible aerial geometry, no unrelated settlements.使い方
- パラメータを含む元のプロンプト全体をコピーします。
- 被写体、文字、構図を自分の要件に置き換えます。
- 対応する外部ツールで使用し、利用条件と結果を確認します。
開始前の確認
- 参考画像の要否を確認してください。互換モデルの表示は、そのモデルでの生成実績を意味しません。
- 名前、数値、画像内の文字を確認します。モデルの版で結果は変わります。





