KAGE-SAN IS NOT HERE / EP1. p13

The recipe

Prompt

Use GPT-5.4 Image 2. Make one picture: a single vertical 2:3 manga page. STYLE: Full-color manga page. A stylish fusion of shoujo manga and American comic book art, with strong ninja aesthetics. Characters: tall, slender, elegant proportions; large expressive glossy eyes with detailed highlights; fine flowing hair with sharp highlights. Inking: bold confident outlines, heavy spot blacks, dramatic chiaroscuro and rim lighting, Ben-Day halftone dot shading. Ninja motifs: long crimson scarves, crescent-moon clan crest, shuriken, smoke bursts, ink-wash brushstroke effects. Palette: midnight navy, ink black, steel grey, crimson accent. Layout: cinematic panel composition, clean white gutters, Western LEFT-TO-RIGHT reading order. Lettering: clean English comic lettering in white speech balloons, all-caps comic font, short lines, perfectly spelled. Keep every character exactly consistent with the attached character reference sheets. [ROMANCE MODE] Shoujo-leaning rendering: soft bloom lighting, sparkles, floating petals and flower screentones, gentle pink highlights, delicate eye shine, blushing cheeks. [ACTION MODE] American-comic-leaning rendering: dynamic foreshortening, speed lines, heavy shadows, halftone dots, impact frames, bold sound effects in English (WHOOSH, CLANG, THUD). PAGE 13 of 16. Portrait 2:3. 3 panels, left-to-right reading order. Panel 1 (huge splash, top three-quarters): mid-air in the spiral stairwell, KAGE-SAN (shinobi mode) catching HIKARI in a princess carry, his crimson scarf spiraling around them both. Moonlight pours through a tall stairwell window behind them. Shoujo sparkles and petals mixed with bold comic halftone shadows. HIKARI's eyes widening, staring up at his masked face; only his storm-grey eyes and silver streak are visible. Panel 2 (bottom left, close-up): HIKARI's flushed face, eyes shimmering. HIKARI: "Who... are you?" Panel 3 (bottom right, close-up): KAGE-SAN's eyes above the mask, gentle for one instant. KAGE-SAN: "Just a shadow." Avoid: multiple images or variations, image grids, Japanese or garbled text, misspelled words, right-to-left panel order, extra fingers or limbs, inconsistent character designs, watermarks, signatures, photorealism, gore.

Model

GPT-5.4 Image 2

Cost

$0.21