
原文
GPT Image 2 · en
Scene:
Create a square competition-style pastry studio photograph with one sculptural chocolate showpiece on a dark pedestal against a controlled neutral or black backdrop. Softbox reflections should skim the tempered surfaces with glass-like sharpness, while shadows stay deep and elegant. The object must feel large, balanced, and professionally lit, as if documented after winning a grand pastry prize. Use a square dark tabletop or neutral studio background with controlled light and enough breathing room for multiple components to float or rest in separated order. The composition should feel analytical and premium.
Subject:
Lay out or suspend the components of a showpiece in exploded order: curved chocolate arcs, slender supports, a base, amber sugar inserts, and a few metallic accents. The parts should feel engineered for assembly, like a pastry equivalent of an exploded product diagram but without text. The image must remain elegant and tactile.
Important details:
Tempered chocolate realism is critical: mirror reflections, crisp edges, matte-to-gloss transitions, gold leaf, sugar accents, and structural balance all need to feel plausible. The sculpture can be impossible in composition, but never sloppy in craft. Every support point and reflection line should suggest deliberate pastry engineering. Every component needs a credible structural purpose and a distinct material role. Reflection lines and cast shadows should help explain assembly logic without any labels.
Use case:
Chocolate lacquer sculpture single study 8/9, revealing the structural logic behind a competition showpiece.
Constraints:
No readable text, no watermark, no logos, preserve exact 1:1 aspect ratio, no candy-bar packaging cues, no readable competition banners, no cartoon dessert palette, no smudged fingerprints, no cluttered workshop background, no technical text callouts, no workshop mess, no plastic support rods.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。