neroidea
由 ChatGPT 根据所示提示词生成,点击图片可全屏查看。

档案卡现场手绘记录风上下对比图

ChatGPT海报比例 3:4
提示词来自 @Hamburgerai

Use case: style-transfer Asset type: Vertical top-bottom comparison style transfer poster Primary request: Transform any input real photo into a "field hand-drawn archive card record style" top-bottom comparison transfer image. Input images: The input photo is the sole source of truth, using only the subject, pose, quantity, structure, spatial relationships, and confirmable colors from the photo. Scene/backdrop: The image is vertical with approximately a 3:4 ratio; the upper half completely retains the real photographic texture, original composition, and natural light of the input photo as a clear photo recording area; the lower half becomes an off-white, vintage archive card with slight fibers and old paper deckle edges. Subject: No limit on subjects, can be people, animals, buildings, street scenes, nature, vehicles, interiors, food, or common objects; extract the most recognizable subject and two or three details from the original photo, do not fix the theme to ships or ports. Style/medium: The lower half uses dark brown pen/marker lines, slight dry ink scratches, and a small amount of crayon or vintage screen-printed color blocks, like travel archive cards, field survey cards, or hand-drawn observation records on the back of old postcards; lines retain hand-drawn jitter, local repeated strokes, and uneven ink, do not make it a smooth vector illustration or a common filter. Composition/framing: The lower half reorganizes the original photo information and does not copy the original framing proportionally: scale down the subject or magnify parts into archive illustrations, arrange a small number of local detail boxes, arrows, numbers, legends, or material fields around them; leave clear off-white breathing space, the layout looks like a complete physical card. A short English title, a short field related to the photo facts, and an extremely short observation can be added, but no long paragraphs of text. Lighting/mood: The upper half is natural and real; the lower half has soft paper-surface light, a restrained vintage record feel, quiet and observational. Color palette: Based on off-white, dark brown, grey-blue, or dark red, extracting only one or two low-saturation accent colors from the original photo; do not apply a fixed yellow-blue color scheme. Materials/textures: Visible paper fibers, old card edges, slight printing misalignment, dry ink broken lines, handwritten fields, tiny numbers, and local color blocks; textures must participate in the composition, not just adding a grain filter. Text (verbatim): Only generate short titles/fields/observation sentences explicitly specified by the prompt; if the location cannot be confirmed, do not fabricate place names, dates, coordinates, brands, URLs, or long paragraphs. Constraints: The upper half must be a recognizable real original image; the lower half must be a hand-drawn archive reconstruction of the same photo content; strictly top-bottom partitioning, no side-by-side, no nine-grid, no multiple different photos, no "Original/AI" labels, no watermarks; maintain subject identity, quantity, pose, architectural structure, topographic relationship, and observation angle; the lower half must change at least two of: subject scale, cropping, information density, or spatial organization; the lower half must not be a mere isomorphic copy of the original image. Avoid: Pure color grading, common photo filters, 3D rendering, cartoon templates, irrelevant travel icons, adding people or landmarks out of thin air, fake Chinese, QR codes, long paragraphs, or over-crowded sticker collages.

提示词解读

想象一张纵向画布被严格切成上下两半。顶部是一张清晰、逼真的照片,体现了主体的原始色彩与自然光照。底部则变成一张现场手绘档案卡:淡米白色纸张表面可见纤维、轻度磨损的旧纸边缘以及暗棕色笔线的跳跃笔画,配合少量与原图相近的低饱和色块。 划分方式严格:上半部保持原始摄影细节,下半部则以手绘方式重构,同一主体的比例和空间关系可能会被缩放或者近大后再绘制细节盒子、箭头与小号注释。配色仅提取一到两种低饱和度的原图突出色,以营造复古档案卡的质感,整体构图关注纸张纹理与手绘痕迹。 使用该提示时,只需保留上下对比的结构和对主体完整性的坚持。若想换成水彩素描或墨迹渲染等,亦可将“手绘档案卡”改为相应媒介;同时维持顶半透明清晰、底半手绘约定的字句与图案排布,避免大量文字或数字标签。这样既能保持原有效果,又能灵活演绎多种复古风格。

#poster#opennana

觉得自己能猜中这条 prompt?来玩 Promptle