oriental-editorial-poster
Create finished Chinese cultural, fantasy, and minimalist text-material editorial posters or publication covers with mandatory Image 2 style image generation. Use when a user gives a title, theme, exhibition/brand/cultural content, reference images, short keyword batches, ABC tests, or asks for 文字创意极简, fantasy text-driven visuals, 东方编辑海报, art-book covers, refined sparse poster tests, or culturally grounded Chinese typography posters. Supports v2 title refinement, real cultural evidence, Mode A direct creative image posters, Mode B steadier complete image-generated poster layouts, and A/B/C comparison batches; avoid generic AI Chinese-style posters.
dacnay816y62-hub · 已有 AI 审核记录
安装与使用
运行命令前先阅读来源与依赖说明。复制命令不会自动安装或执行。
npx skills add dacnay816y62-hub/fantasy-dongfang-jianyuehaibao --skill oriental-editorial-poster技能说明
name: oriental-editorial-poster description: "Create finished Chinese cultural, fantasy, and minimalist text-material editorial posters or publication covers with mandatory Image 2 style image generation. Use when a user gives a title, theme, exhibition/brand/cultural content, reference images, short keyword batches, ABC tests, or asks for 文字创意极简, fantasy text-driven visuals, 东方编辑海报, art-book covers, refined sparse poster tests, or culturally grounded Chinese typography posters. Supports v2 title refinement, real cultural evidence, Mode A direct creative image posters, Mode B steadier complete image-generated poster layouts, and A/B/C comparison batches; avoid generic AI Chinese-style posters."
东方文化编辑海报 v2
Treat this as a research-and-composition skill with mandatory image generation, not a generic one-prompt “Chinese-style poster” shortcut.
The goal is a contemporary editorial page that uses cultural evidence and material logic: real paintings, historical prints, artifacts, archival photography, calligraphy samples, book pages, material details, or generated reinterpretations that are clearly not presented as archival fact. In Mode A, image generation may create the whole poster with a more expressive image-title fusion. In Mode B, image generation still creates the whole poster, including title placement and typographic composition, but with a calmer publication-cover balance between text field, image evidence, and layout hierarchy.
Read editorial-systems.md before composing.
For short keyword tests, quick batches, "fantasy", "文字创意极简", or outputs intended to match the user-approved sparse text-material examples, also read fantasy-text-minimal.md. That overlay is the default taste layer for fast 1-6 keyword poster batches: one strong material evidence field, one semantic title action, restrained text, one useful interruptor, no fake small text, and no generic Chinese decoration.
When judging taste or running ABC comparisons, inspect v2-tests/selected-triptychs/selected-overview.jpg first. It captures the current approved direction: refined short Chinese titles, a single cultural/material proof, one image-title relationship, and calm editorial space.
v2 defaults from Image 2 testing
These rules supersede earlier broad "fantasy oriental" tendencies.
- Use Image 2 / image generation for actual poster tests; do not stop at prompt writing when the user asks for output images.
- When the user is testing Chinese text fidelity, asks to fix question marks/typos, or explicitly rejects post-added text, the visible H1 must be generated inside the image model pass. Do not add, replace, or correct the main title later with PIL, SVG, HTML/CSS, Photoshop, or any other layout fallback. If the title is wrong, simplify the prompt/title and regenerate.
- Prefer real, recognizable cultural evidence over invented cultural mashups: porcelain, silverwork, lacquer, garden windows, opera costume details, shadow puppets, city wall, grotto stone, archive maps, fabric, paper, waterline, patina, or craft marks.
- Refine the main Chinese title before prompting. The H1 should usually be 1-4 Chinese characters. Carry the longer topic through evidence, subtitle, English tag, or small copy.
- Do not over-explain with the title. For example, "徽州天井" can become "天井" when Huizhou is visible through architecture; "景德镇青花" should emphasize "青花瓷" or "青花" rather than forcing every proper noun into the H1.
- For abstract topics such as 乡愁, 婚姻, or 城市缩影, choose a tangible carrier first: threshold, ring, ledger, map fold, ticket, old wall, road trace, household object, or reflected surface.
- For place names such as 北京 or 山西大同, avoid landmark collage. Use one local material clue: hutong brick, map edge, wall section, grotto stone, roof ridge, sign shadow, or archival street fragment.
- Favor finished editorial posters over tourism posters, museum placards, guochao ecommerce graphics, cyberpunk wallpaper, or pure mood images.
- In batches, evaluate by complete groups, not isolated favorites. A good group has a coherent subject and three clearly different treatments.
ABC test modes
Use A/B/C only when the user asks for tests, comparisons, or multiple directions.
| Mode | Name | Output logic |
|---|---|---|
| A | 古图档案介入 | Use archive, rubbing, map, print, old paper, or historical-image logic. The title behaves like an inscription, specimen, crop, or archival interruption. |
| B | 单一物证 | Use one object or material as the visual boss. The title is restrained and the object carries the cultural identity. |
| C | 字体结构实验 | Let the title become a structure, mask, crop-window, relief, section, or frame. Keep it readable and avoid turning the whole subject into a literal giant glyph. |
For a five-topic ABC test, produce 15 images and save matching prompts. Also make a contact sheet or triptych view when the user needs review.
Poster-first correction defaults
These defaults override any earlier tendency to make a single clever image.
- Treat the deliverable as a finished poster, cover, or publication page. The result must have layout hierarchy, real title placement, readable metadata rhythm, and poster-level composition. Do not let the generated image become the whole answer unless the user explicitly asks for a creative-image experiment.
- Decorative hairlines, dots, registration marks, circles, rulers, and random editorial ornaments are optional, not mandatory. Add them only when they clarify hierarchy, alignment, scale, or source logic. If they do not have a job, omit them.
- For intangible cultural heritage, city culture, regional identity, exhibitions, festivals, craft subjects, and any multi-character title or topic, default to Mode B: image-generated poster layout unless the user explicitly requests Mode A.
- In Mode B, the image model should generate the complete poster in one image-generation step, includin