结果比我预期的更好。一个用于 GPT Image 2 的漫画/动漫提示：

This turned out better than I expected. A manga/anime prompt for GPT Image 2: INPUT ::= [TOPIC / TITLE / SUBJECT] TITLE_TOKEN ::= derive_short_display_title(INPUT) HERO_FORM ::= infer_most_iconic_visual_embodiment(INPUT) SECONDARY_FORMS ::= infer_supporting_shapes_symbols(INPUT) ACCENT_COLOR ::= infer_single_semantically_relevant_accent(INPUT) BASE_PALETTE ::= [ink_black, warm_paper_white, ACCENT_COLOR] PAPER_DNA ::= layered cardstock / folded paper / cut-paper relief / sharp polygon facets CANVAS ::= +--------------------------------------------------------------+ | [TITLE typography dominates left/center] | [HERO_FORM] | | | integrated into | | geometric shards / rules / fragments | paper architecture| |--------------------------------------------------------------| | lower framing motifs / cropped shapes / visual ballast | +--------------------------------------------------------------+ TYPOGRAPHY_RULES ::= - construct TITLE_TOKEN from custom angular paper-cut letterforms - letters should feel physically assembled from layered cardstock - use sharp wedges, diagonal cuts, clipped terminals, inset shadows - selectively fracture glyphs into geometric extensions that point toward HERO_FORM - preserve high title readability despite abstraction HERO_RULES ::= - translate HERO_FORM into faceted folded-paper sculpture - build anatomy/object geometry from planar shards rather than smooth surfaces - integrate silhouette into the same graphic system as typography - allow fragments from HERO_FORM to invade title space and vice versa - use ACCENT_COLOR only in strategic interior planes or semantic focal regions LIGHTING ::= shallow-relief studio light + crisp paper-edge shadows + tactile thickness + subtle elevation from background STYLE := editorial poster × papercut sculpture × geometric constructivism × luxury title design × restrained negative space FORBID := hardcoded subject generic font smooth 3D plastic rainbow palette separate illustration beside text busy realistic background

结果比我预期的更好。一个用于 GPT Image 2 的漫画/动漫提示： INPUT ::= [TOPIC / TITLE / SUBJECT] TITLE_TOKEN ::= derive_short_display_title(INPUT) HERO_FORM ::= infer_most_iconic_visual_embodiment(INPUT) SECONDARY_FORMS ::= infer_supporting_shapes_symbols(INPUT) ACCENT_COLOR ::= infer_single_semantically_relevant_accent(INPUT) BASE_PALETTE ::= [ink_black, warm_paper_white, ACCENT_COLOR] PAPER_DNA ::= layered cardstock / folded paper / cut-paper relief / sharp polygon facets CANVAS ::= +--------------------------------------------------------------+ | [TITLE typography dominates left/center] | [HERO_FORM] | | | integrated into | | geometric shards / rules / fragments | paper architecture| |--------------------------------------------------------------| | lower framing motifs / cropped shapes / visual ballast | +--------------------------------------------------------------+ TYPOGRAPHY_RULES ::= - construct TITLE_TOKEN from custom angular paper-cut letterforms - letters should feel physically assembled from layered cardstock - use sharp wedges, diagonal cuts, clipped terminals, inset shadows - selectively fracture glyphs into geometric extensions that point toward HERO_FORM - preserve high title readability despite abstraction HERO_RULES ::= - translate HERO_FORM into faceted folded-paper sculpture - build anatomy/object geometry from planar shards rather than smooth surfaces - integrate silhouette into the same graphic system as typography - allow fragments from HERO_FORM to invade title space and vice versa - use ACCENT_COLOR only in strategic interior planes or semantic focal regions LIGHTING ::= shallow-relief studio light + crisp paper-edge shadows + tactile thickness + subtle elevation from background STYLE := editorial poster × papercut sculpture × geometric constructivism × luxury title design × restrained negative space FORBID := hardcoded subject generic font smooth 3D plastic rainbow palette separate illustration beside text busy realistic background