GPT Image 2.5 makes a six-person poster thumbnail with percentage-based layout
The author used a prompt with percentage-based placement for each figure to build a six-person medieval court ensemble image, and compares it with earlier group thumbnails to note better composition control.
#AIイラスト 昨日のイラストを元にサムネイル用集合絵完成 画像2~4は過去作の5人以上集合絵サムネイルだが、以前よりは構図の制御に慣れてきた気がする しかし、ルネティア以外を赤系統で統一したのは一応理由あるけど、画面が派手 ちなみに使用モデルはGPT Image 2.5 【使用プロンプトメモ】 A masterpiece. Top quality. A group portrait of six people. Set in the world of a fictional medieval court drama. The background features red and yellow as the primary colors. Blur the boundaries or disguise them with lines and patterns. Do not include text, but frame the image with a thin border to give it a poster-like look. Composition: Top left: A close-up of the woman’s profile from Reference Image 1. It occupies approximately 60% of the total height and 55% of the total width of the frame. Bottom left: The two men standing side by side from Reference Image 2. They occupy approximately 40% of the total height and 45% of the total width of the frame. Bottom right: The woman holding a book from Reference Image 4. She occupies approximately 55% of the total height and 55% of the total width of the frame. Top right: Reference Image 3 shows two people standing side by side. They are depicted slightly closer to the viewer than the two men in the bottom left. They occupy approximately 45% of the total height and 45% of the total width of the frame.
This post quotes another post: x.com/kurohachi05/status/2104498629763637411