UI はローカライズ済みです。スキルの紹介と説明文書は選択した言語へ AI が自動翻訳し、コード・コマンド・パス・専門用語は原文のままです。原作者の文書が優先されます。
スキルを探す
デザイン · SKILL.md

folk-ink-landmark-portrait

folk-ink-landmark-portrait

Turn travel photos with recognizable architecture or human-made landmarks into 3:4 folk ink-and-color posters, with either a photo comparison layout or a full-canvas redraw. Do not use when no landmark is recognizable.

Independent GitHub snapshot · Automated classification · Original documentation governs

スキルの説明

GitHub

name: folk-ink-landmark-portrait description: "Turn travel photos with recognizable architecture or human-made landmarks into 3:4 folk ink-and-color posters, with either a photo comparison layout or a full-canvas redraw. Do not use when no landmark is recognizable."

Folk Ink Landmark Portrait

Delivery and mode selection

  • Use only for photos with clear architecture or a human-made landmark. If none is recognizable, do not invent one; explain that the image is unsuitable.
  • Produce one standalone 3:4 vertical poster per source image. Never combine photos.
  • Offer two modes when the user has not already chosen:
    1. Split comparison — real source photo above, folk ink portrait below.
    2. Full canvas — the whole poster becomes the folk ink portrait.
  • Use one mode for all uploaded photos unless the user maps modes per image.
  • Read only the selected mode guide: split-comparison.md or full-canvas.md.
  • Return a ready-to-use prompt only when explicitly requested; otherwise use the available image generation or editing capability.

Shared visual rules

Preserve landmark identity, count, viewpoint, height-width proportion, perspective, and defining relationships among roof, dome, tower, opening, steps, columns, sculpture, entrance, base, bridge, wall, or shoreline. Reduce nonessential tourists, vehicles, signage, repeated windows, masonry seams, ornament, dense plants, and photo noise only in the illustrated region. Never remove defining structure or add regional symbols absent from the source.

On matte warm-white paper, use heavy slightly hesitant black brush or marker lines: variable weight, local retracing, occasional breaks, small ink blocks, and a few open contours. Structure may feel handmade but must remain accurate. Add only 2–3 signature colors extracted from the source as matte gouache blocks with visible brush marks, slight overrun, misregistration, gaps, and exposed paper.

Keep generous continuous paper white. Before rendering, lock readable text from source facts only: a 1–3-word English title, STUDY NO. XX or LANDMARK NOTE, and a 5–10-word observation. If the place is unknown, use a truthful scene theme; never guess a name, date, or coordinates.

Avoid

Fixed regional templates, invented local symbols, calligraphy or fake characters, copied architecture, detailed technical pen drawing, automatic photo outlines, vectors, closed digital fills, transparent watercolor, halftones, full backgrounds, excessive detail, unreadable text, signatures, handles, logos, QR codes, URLs, and watermarks.

保存したスキル