OpenZhenZhen AI 프롬프트 갤러리 · gpt-image-2.5 · image

기하학 동반자 안내서 상하 비교 포스터

Geometric Companion Guide: Split-Screen Comparison Poster

‘기하학 동반자 안내서 상하 비교 포스터’은(는) gpt-image-2.5로 만든 작품입니다. OpenZhenZhen에서 원본 프롬프트와 생성 예시를 살펴보고 새로운 창작에 활용해 보세요.

기하학 동반자 안내서 상하 비교 포스터 — AI image creation
기하학 동반자 안내서 상하 비교 포스터

프롬프트

중국어 프롬프트

将绑定到 identity_reference（ref-10a-v1）的唯一整幅照片作为完整图像来源和唯一身份参考。照片中是一位成年女性，居中直视镜头的上半身肖像，深棕色中分长发，穿黑色细肩带上衣，背景为纯净青蓝色。创作一张竖版上下对比海报。上部约占全画布45%，下部约占55%；使用清晰自然的水平分界。禁止左右并排、网格或多图拼贴。 上部——原始照片 将实际输入照片以高保真摄影方式置于上部。保留同一女性及可辨认身份、单人数量、可见姿势、直视方向、裁切、光线方向、色彩和空间关系。不得重绘、模糊、重构或修饰成另一个人，也不得替换照片任何部分。 下部——几何伴游说明页 将同一女性重绘成克制的纸面观察说明页，不要做成照片或普通卡通滤镜。改变主体尺度与信息密度，而非机械描摹照片构图。将可见的头部、头发、面部、肩部和黑色衣物拆解为少量清晰的平面几何形、粗细有变化的黑色轮廓，以及恰好四种从原照提取的色块：青蓝、近黑、暖象牙白和柔和桃色。保留身份、单人数量、直视方向、姿势和最具辨识度的可见结构，让下部一眼仍能认出是同一位女性。底色为温暖米白纸张，带轻微纤维、不均匀印刷和低对比颗粒；主体与标注之间留出呼吸空间。 用箭头、虚线引导线、两个小型编号圆点（01、02）、简洁结构线和局部放大框构成清晰的说明关系。每一条箭头、虚线和局部放大框都必须精准指向原照中真实可见的特征：中分发缝、头发轮廓或黑色肩带布料。不得暗示虚构地点或添加物体。几何拆解清楚简洁，不得变成复杂工程图。不得增加其他人物或无关地标。 可见文字 仅准确呈现以下三条简短英文，各一次：标题“Portrait Study”；一条真实观察字段“Observation: center-parted hair”；以及一句极短指令“Compare the outline.”。观察字段必须符合照片中可见事实。编号圆点01和02仅作标注锚点。不得添加中文、段落、网址、二维码、品牌、水印、“原图/AI图”徽标或其他文字与字母。 避免只调色、只叠加颗粒、照片描摹、下部仍是写实照片、无意义装饰贴纸、固定狗或海滩图案、额外人物，以及密集难读的小字。只输出一张完成的竖版海报：上方原照、下方几何说明页、一道自然水平分界；不得增加其他分区或解释文案。

영어 프롬프트

Use the single full photograph bound to identity_reference (ref-10a-v1) as the complete source image and sole identity reference. It shows one adult woman in a centered, direct-gaze upper-body portrait, with long center-parted dark-brown hair, a black thin-strap top, and a plain teal-blue background. Create one vertical up-down comparison poster. Keep the upper region about 45% of the canvas and the lower region about 55%; use one clear, natural horizontal boundary. No side-by-side layout, grid, or multi-image collage. UPPER — UNALTERED PHOTO Place the actual input photograph in the upper region as high-fidelity photography. Preserve the same woman, recognizable identity, one-person count, visible pose, direct gaze, crop, lighting direction, colors, and spatial relationships. Do not redraw, blur, reconstruct, retouch into another person, or replace any part of the photo. LOWER — GEOMETRIC COMPANION MANUAL Redraw the same woman as a restrained printed-paper observation page, not as a photo or ordinary cartoon filter. Change scale and information density rather than mechanically tracing the photo composition. Deconstruct her visible head, hair, face, shoulders, and black garment into a few clean flat geometric planes, varied-weight black outlines, and exactly four source-derived color blocks: teal-blue, near-black, warm ivory, and muted peach. Preserve identity, one-person count, direct-gaze direction, pose, and the most distinctive visible structure so the lower drawing is immediately recognizable as the same woman. Use a warm off-white paper ground with subtle fibers, slight uneven printing, and low-contrast grain; leave breathing room around the subject and annotations. Build a clear explanatory relation with arrows, dotted guide lines, two small numbered dots (01 and 02), concise structural strokes, and partial enlargement boxes. Every arrow, dotted line, and enlargement box must point precisely to a feature actually visible in the source: the center hair part, the hair silhouette, or the black strap fabric. Do not imply a location or add objects. Keep geometry legible and simple, never an intricate engineering diagram. No extra figures or irrelevant landmarks. VISIBLE WORDING Render exactly these three short English strings, each once: title “Portrait Study”; one truthful observation field “Observation: center-parted hair”; and one very short instruction “Compare the outline.” The observation must correspond to the visible photo. The numbered dots 01 and 02 are callout markers only. Add no Chinese, paragraph, URL, QR code, brand, watermark, badge saying original/AI image, or any other words or lettering. Avoid color-only changes, grain-only overlays, photo tracing, a realistic photo in the lower region, meaningless decorative stickers, fixed dog or beach motifs, extra people, and dense or unreadably small text. Output exactly one finished vertical poster with the top photo and lower geometric manual, one natural horizontal division, and no other panel or explanatory copy.

API 모델 보기

더 많은 이미지·동영상 프롬프트 보기 · 이미지 생성 · 동영상 생성 · AI 캔버스에서 만들기