墙绘分身:现实与手绘的对话 · 案例 11 / 6
人像 · 摄影 · 插画

墙绘分身:现实与手绘的对话

将照片中的真实人物与身后墙上的手绘黑白漫画风肖像并置,形成一实一绘的趣味对照,保留原始摄影质感与身份细节。

GPT-Image-2.5写实漫画风黑白手绘人像概念艺术

完整提示词

The uploaded photograph is the sole source of truth for the people, identities, clothing, accessories, poses, camera framing, and scene content. Create one seamless, photorealistic image in which every primary real subject appears together with a corresponding hand-drawn mural portrait painted directly onto a wall behind or beside them. PRIMARY-SUBJECT SELECTION Identify only the primary subjects: the people who are clearly the intended photographic subjects, compositional focus, and action focus of the uploaded image. Do not treat background pedestrians, distant figures, reflections, people in posters or screens, statues, or accidental bystanders as primary subjects. Preserve incidental background people when appropriate, but never create mural counterparts for them. ONE-TO-ONE SUBJECT COUNT If the source photograph contains N primary subjects, the final image must contain exactly: - the same N real primary subjects; and - exactly N mural counterparts. Every real subject must have one—and only one—clearly corresponding mural version. Never add, remove, merge, split, duplicate, replace, or swap any primary subject. For group photos, maintain the original left-to-right identity order and make every pairing visually unambiguous. LOCK THE REAL SUBJECTS Treat all real primary subjects as protected, identity-locked regions. Preserve them as closely as possible to the uploaded photograph. Strictly preserve each real subject’s: - facial identity and recognizable facial structure; - apparent age and natural skin appearance; - hairstyle, hair color, and hair accessories; - expression, gaze direction, and head angle; - exact clothing, garment colors, prints, patterns, layers, and visible details; - glasses, jewelry, hats, bags, footwear, and other accessories; - body proportions and visible anatomy; - arm directions, hand gestures, finger placement, leg positions, body orientation, and weight distribution; - relationship with all held, worn, touched, or supporting objects; - original photographic realism. Do not beautify, face-swap, age-shift, reshape, restyle, recolor, change clothing, change expression, alter the pose, or convert any real person into an illustration. Do not mirror or reverse the image. Do not invent body parts hidden or cropped outside the source frame. PRESERVE THE PHOTOGRAPH Keep the original camera angle, perspective, lens character, shot type, crop, aspect ratio, subject scale, depth of field, lighting direction, shadows, color atmosphere, and overall environment. Edit the background only where necessary to create a suitable mural surface. If the source already contains a usable wall, preserve that wall and integrate the drawing into it. If no suitable wall exists, minimally convert only an adjacent background surface into a believable light-colored or naturally textured wall. Do not expand the frame, zoom out, reconstruct off-frame content, or redesign the entire location. MURAL COUNTERPARTS Place one large two-dimensional mural portrait behind or beside each corresponding real subject. The mural should usually be approximately life-size to 1.6 times the real subject’s apparent size, depending on the available wall space. Each mural must strictly reproduce its corresponding subject’s: - recognizable identity; - face shape and visible facial characteristics; - hairstyle and accessories; - exact outfit design, silhouette, layers, patterns, and major details; - expression and gaze direction; - head angle; - arm directions and hand gestures; - leg positions; - body orientation, balance, and complete pose; - interaction with important props or objects. The mural may translate fine details into hand-drawn marks, but it must not reinterpret, exaggerate, simplify away, mirror, or change the meaning of the pose. Do not give the mural a different outfit, hairstyle, expression, gesture, prop, or body position. For multiple subjects, keep every mural spatially close to its own real counterpart. Never mix one person’s face with another person’s clothing, pose, accessories, or mural. MURAL DRAWING STYLE Render every mural as a handcrafted monochrome wall illustration combining: - expressive black ink linework; - graphite and pencil-sketch texture; - Japanese manga-inspired editorial drawing; - casual urban doodle aesthetics; - confident thick-to-thin contour lines; - loose construction strokes; - irregular handmade marks; - selective cross-hatching and scribbled shading; - a few solid-black accents; - simplified tonal modeling; - recognizable portrait likeness. Use the physical wall color as the drawing’s base, allowing the wall surface to remain visible through unfilled areas. The mural must remain clearly two-dimensional and visibly hand-drawn—not photorealistic, not fully colored, and not a second real person. Preserve important clothing patterns and accessories through simplified but recognizable linework. Give faces enough detail to identify the correct person while maintaining the loose manga-sketch character. PHYSICAL WALL INTEGRATION The mural must appear genuinely drawn or painted onto the architecture. Make all linework conform naturally to: - the wall’s plane and perspective; - surface grain and roughness; - cracks, seams, panels, bricks, or worn areas; - existing light and shadow; - architectural edges and partial occlusion. The wall texture should subtly interrupt or absorb the ink and pencil marks. The mural must not appear as a floating graphic, digital overlay, sticker, framed artwork, paper poster, banner, projection, or separate canvas. COMPOSITIONAL RELATIONSHIP Create a playful visual dialogue between each real person and their wall-drawn counterpart. Position the murals so the matching identity, outfit, and pose are immediately readable, while keeping the real subjects visually dominant and unobstructed. Do not cover faces, hands, key accessories, or important pose details. Preserve believable spatial depth and natural occlusion. The final composition should feel like a real photograph captured at an actual location—not a studio composite. DOODLES AND HANDWRITTEN TEXT Add only a few small black hand-drawn accents around the mural, such as hearts, crowns, stars, sparkles, question marks, exclamation marks, smiles, arrows, or short emphasis lines. Keep all decorations sparse, casual, and secondary to the people. Optionally add one short, context-appropriate English handwritten phrase containing 2–5 simple words. It may refer to the subject’s mood, pose, clothing, activity, or location. The phrase must be correctly spelled, clearly readable, and visually integrated into the same wall-drawing style. If reliable spelling cannot be produced, omit the phrase instead of generating broken or meaningless text. Do not add signatures, artist names, social-media handles, logos, labels, long sentences, random letters, or watermarks. FINAL APPEARANCE The finished result must look like a single authentic lifestyle photograph of real people posing beside custom wall murals of themselves. The people must remain unmistakably real and faithful to the uploaded photograph. Their mural counterparts must be unmistakably flat, monochrome, hand-drawn wall art with strict one-to-one identity, outfit, expression, and pose correspondence. From a distance, the real subjects and their mural counterparts should form a clear paired composition. At close range, the wall texture, irregular ink strokes, pencil marks, and cross-hatching should remain visible and physically convincing. PRIORITY ORDER When any instruction competes for space or clarity, follow this priority: 1. Preserve the exact primary-subject count and one-to-one correspondence. 2. Preserve the real subjects’ identities and photographic appearance. 3. Preserve the exact clothing, accessories, expression, pose, and object interactions. 4. Preserve the original crop, camera viewpoint, lighting, and environment. 5. Integrate each mural naturally into a real wall. 6. Add optional decorative doodles and text only when space permits. HARD FAILURES TO AVOID No changed identity. No changed face. No beauty filter. No changed age. No changed hairstyle. No changed clothing or clothing color. No changed expression. No changed gaze. No changed pose or gesture. No mirrored pose. No added, missing, duplicated, merged, or swapped primary subject. No missing or unmatched mural. No identity mixing between subjects. No new person. No mural for a background bystander. No illustrated real subject. No photorealistic mural subject. No full-image illustration. No detached or floating drawing. No sticker, frame, poster, paper, projection, or separate canvas. No collage, diptych, split screen, comparison layout, or poster design. No expanded framing or invented off-frame anatomy. No excessive doodles. No cluttered replacement background. No malformed face, hands, fingers, limbs, or object interactions. No garbled text. No logo, signature, social-media handle, or watermark. Return exactly one complete final image with no explanatory text and no alternate versions.

作者:苏乐

使用这个提示词创作

可以按需微调提示词或者上传参考图片,点击生成,查看结果,并及时下载。

(可选)

生成图片

生成期间,请勿关闭页面。如已关闭,可通过右上角的历史记录查看生成进度和结果。

如何使用

  1. 调整提示词

    在创作表单中查看预填的提示词,按需修改主体、场景、颜色或风格。

  2. 上传参考图片和比例

    如果页面提供参考图片上传入口,可以选择上传,也可以留空。请上传拥有使用权的图片,并按需调整图片比例。

  3. 生成图片

    模型已内置,点击生成,稍等片刻,等待生成结果。

  4. 查看并下载

    生成完成后查看结果并下载图片。想尝试其他效果,可以修改提示词或参考图片后再次生成。

使用技巧

上传一张或多人的清晰照片。提示词会严格保留原片中每位主体的面部特征、发型、服装、配饰、姿态及环境光影,并在其身后或身旁的墙面上生成对应的黑白手绘壁画。壁画风格融合墨水线条、铅笔素描肌理与日系漫画速写感,与真实人物形成一对一配对。可替换主体、场景与墙面类型,但需确保原图有可用的墙面空间;若无合适墙面,生成时会自动将部分背景转换为浅色或自然纹理的墙壁。装饰涂鸦和简短英文手写短语为可选元素,仅在空间允许时添加。生成结果需根据实际墙面透视、纹理和光线效果进行微调。

常见问题

适合新手使用吗?

适合。不需要专业 AI 知识,也不需要学习画布操作。选择喜欢的提示词,在页面按需上传参考图或者微调提示词,模型已配置好,点击生成即可。第一次使用也能轻松开始创作。

提示词是免费的吗?

是的。这里的提示词收集整理自公共互联网,可免费查看、复制和使用,也可以按自己的创意修改。也可以通过作者,查看原始来源。

提示词怎么使用?

可以复制提示词,粘贴到 ChatGPT 等支持图片生成的工具中使用;也可以直接使用本页创作表单,模型已配置好,按需修改提示词、上传可选参考图并调整比例,点击生成即可。不同工具或模型的生成效果可能不同。

生成图片是否免费?

可以免费体验!注册即可获得免费体验算力,用于生成图片。体验满意后,可以订阅会员或购买算力,进行更多 AI 图片创作。不同模型和配置消耗的算力不同,具体以生成前的提示为准。

平台有哪些 AI 图像模型?

Oimi 已接入 GPT-Image-2、GPT-Image-2.5、Nano Banana 2.1、Nano Banana Pro、Midjourney V8.2 / V7、Niji Journey(NJ)、Seedream 5、Grok Image 2、万相 2.7 和 Z-Image 等图像模型,满足人像、产品、海报、插画及图片编辑等不同创作需求。无需在多个平台之间来回切换,也无需分别订阅多个平台的会员,在 Oimi 即可使用多种模型,统一使用算力进行创作。本页使用预配置模型,更多模型可在 AI 无限画布中体验。

积分除了生成图片,还能做什么?

Oimi 是一站式 AI 创作平台,算力还可用于 AI 文本、视频、音频、音乐等创作。浏览 OimiTV 精选模板,可以复用热门 AI 工作流,尝试制作社交媒体趋势视频,让图片创意延伸为视频和音乐作品。

支持多语言提示词吗?

支持!大部分 AI 图像模型支持英语、法语、德语、中文、日语、阿拉伯语等常见语言。可以使用自己熟悉的语言描述创意;不同模型对各语言的理解能力有所差异,生成效果也可能不同。

为什么提示词是其他语言?

提示词库保留了原作者使用的语言和表达,方便还原创作思路。你可以直接使用原文,也可以翻译成自己熟悉的语言后再生成。翻译时建议保留主体、构图、风格等关键描述,翻译后的效果可能略有不同。

可以修改这里的提示词吗?

可以。建议先修改主体或场景,再逐步调整颜色、构图、风格和细节。页面上方保留了原始提示词,方便对照和复制。修改创作表单不会改变提示词库中的原文。

必须上传参考图片吗?

不需要。参考图片为可选项,不上传也可以根据文字生成图片。支持参考图的模型会显示上传入口;上传后,生成时会结合参考图片。想恢复纯文字生成,可以移除参考图。请仅上传您拥有使用权的图片;涉及他人肖像时,请确保已获得相应授权。

为什么生成结果和示例图片不完全一样?

示例图片用于展示创作方向。生成结果会受到提示词、模型、参考图片、参数以及生成随机性的影响,即使使用相同提示词,也可能得到不同结果。想更接近示例,可以保留关键风格描述,并逐次调整少量内容。

手机上可以直接使用吗?

可以。直接在当前页面的创作表单中编辑提示词、从相册上传参考图、生成和下载图片,无需进入画布,也无需安装应用。

图像生成失败或出错会怎样?

如果图像生成任务因系统错误而失败,该任务已扣除的算力会自动退回账户。可以通过历史记录查看详情。内容违规等情况不适用自动退款。 内容规范

我的照片会被存储吗?

上传的照片或者生成的图片会临时存储一周左右,方便未及时下载或本地误删除的用户,在历史记录中查看生成记录并下载。一周后就会删除。 隐私政策

我的照片会被用于训练吗?

不会!所有照片仅用于图像生成,平台不会将其用于训练。

如果我是专业人士,可以在 AI 无限画布上使用吗?

可以。对于需要反复打磨作品的专业创作者,无限画布提供了更自由的创作空间:在 Oimi 的同一个工作区里组织素材、连续生成、编辑和对比图片,还能连接图生视频等后续工作流,无需在多个工具之间来回切换。可以通过表单中的画布入口打开,也可以将提示词复制到 AI 无限画布中使用。

可以使用 Agent 协助创作吗?

可以。在 AI 无限画布中使用画布 Agent 助手,用自然语言描述想法,让 AI 协助梳理创作思路、创建文本或生成图片,再结合画布中的素材继续调整和创作。

更多相似灵感