在 AI 绘图中，生成准确的文字一直是个难题，而生成“多人且带有准确文字”的图片更是难上加难。传统的做法是后期 PS，但那样不仅麻烦，光影也容易假。今天分享的这套**“群体 T 恤信息传递矩阵 (V4.3)”**，用一种极其巧妙的“物理逻辑”解决了这个问题：把长句子拆解给每个人，一人负责一个字。
核心特征说明
🔠 分布式渲染：一人一字（One Person, One Character），极大降低 AI 拼写错误率，想写什么就写什么。 🛡️ 双层安全：撩开外衣露白 T，这种“千层套路”既有视觉惊喜，又绝对安全合规。 👘 正坐构图：采用 Seiza（日式正坐）+ 弧形排列，解决多人站立时的腿部畸变问题。 👯♀️ 拒绝克隆：强制执行“面部多样性”指令，确保每个人都有独特的发型和表情。
关键技术拆解
🔑 构图的智慧：Semi-circle Arc (半圆弧) 多人合照最怕什么？两边的人脸变形。这套 Prompt 强制使用 gentle, welcoming semi-circle arc（柔和的半圆弧）。这种构图让每个人物到镜头的距离基本相等，配合广角镜头，不仅消除了畸变，还增强了画面的包围感和亲切感。
🔑 动作的张力：The Reveal Action 静态的 T 恤展示略显呆板。我们设计了 playfully lifting the bottom hem（俏皮地撩起底边）这一动作。这不仅增加画面的动态张力，还通过“外衣+内搭”的层次感，丰富了视觉细节。最重要的是，它明确界定了 fully clothed layers，在性感与安全之间找到了完美的平衡点。
🔑 文字的控制：Single Character Logic AI 处理单词容易出错，但处理单个字母/汉字却很擅长。On each white T-shirt is a SINGLE character 这一指令，将复杂的文本生成任务降维打击为简单的图形匹配任务，成功率提升 90% 以上。
🔑 扩展玩法：性别与动态 这套模板的兼容性极强：
-
兄弟会版本：将
women改为men，服装改为jeans and hoodies，瞬间变成热血兄弟连。 -
毕业季版本：将坐姿
Seiza style改为Jumping in the air，配合学士服，就是一张完美的毕业大合照。
实际使用场景
这套系统非常适合需要“人味”和“信息量”并存的场景：
场景一：粉丝应援/打Call。拼出偶像的名字或“生日快乐”，制作独一无二的应援海报。 场景二：创意求婚/表白。生成一组朋友拿着“WILL YOU MARRY ME”的照片，用于求婚视频素材。 场景三：企业文化展示。拼出公司价值观关键词（如 TEAM, GROW），用于 PPT 或文化墙。
通过这套 Prompt，我们不仅解决了 AI 的技术痛点，更创造了一种全新的视觉语汇。以下是完整的 JSON 协议，复制它，修改 text_content，去生成属于你的创意矩阵吧。
| { “template_name”: “Group_Message_TShirt_Reveal_Gen_v4.3”, “template_version”: “4.3”, “template_purpose”: “Generates a photorealistic group photo of diverse women in a semi-circle Seiza pose, revealing text on shirts. Features attractive styling with short skirts while maintaining strict safety standards.”, “applicable_models”: [ “Midjourney v6”, “DALL-E 3”, “Stable Diffusion XL”, “Google Gemini NanoBananaPro” ], “input_assumptions”: “User provides text and count. System organizes subjects in a curved group formation. Skirts are styled shorter/attractive but strictly non-explicit.”, “editable_fields”: [ { “field_key”: “text_content”, “label_cn”: “T恤上的文字内容”, “description_cn”: “想要展示的完整句子或单词（每个女孩衣服上写一个字）。”, “example_values”: [ “生日快乐”, “我喜欢你”, “WELCOME”, “恭喜发财”, “TEAMWORK” ] }, { “field_key”: “text_length”, “label_cn”: “文字字数（人数）”, “description_cn”: “文字内容的字符数量，这将直接决定画面中女孩的人数。”, “example_values”: [ “4”, “7”, “10”, “5” ] }, { “field_key”: “character_ethnicity”, “label_cn”: “人物特征/国籍”, “description_cn”: “定义人物的种族或国籍特征，若不填默认建议使用 Japanese。”, “example_values”: [ “Japanese”, “Chinese”, “American”, “Korean”, “Mixed race” ] }, { “field_key”: “outer_garment”, “label_cn”: “外衣款式”, “description_cn”: “被撩起的上衣款式。默认为黑色薄款毛衣。”, “example_values”: [ “black thin knit sweater”, “school uniform blazer”, “red hoodie”, “loose white shirt”, “blue sports jacket” ] }, { “field_key”: “environment_setting”, “label_cn”: “环境背景”, “description_cn”: “拍摄所在的场景。默认为温暖灯光的室内。”, “example_values”: [ “traditional Japanese tatami room with warm lighting”, “bright modern classroom”, “sunny outdoor grassy field”, “clean white photography studio”, “spacious wooden floor dance studio” ] }, { “field_key”: “aspect_ratio”, “label_cn”: “图片比例”, “description_cn”: “图片的宽才比，建议根据人数多少调整（人多用宽屏）。”, “example_values”: [ “–ar 16:9”, “–ar 3:2”, “–ar 4:3”, “–ar 1:1” ] } ], “generation_constraints”: { “quality”: [ “Photorealistic”, “8k resolution”, “Sharp focus”, “High detail”, “Facial uniqueness” ], “style”: [ “Group photo”, “Seiza sitting pose”, “Semi-circle composition”, “Handwritten marker style text”, “Attractive but safe” ], “negative”: [ “upskirt”, “crotch shot”, “panties”, “sexual”, “nsfw”, “nudity”, “bare midriff”, “stomach skin”, “navel”, “underwear”, “bikini”, “cloned faces”, “same face”, “digital font”, “printed text”, “blur”, “distortion” ] }, “final_image_prompt”: “A wide-angle, photorealistic group photograph of {text_length} young {character_ethnicity} women arranged in a **gentle, welcoming semi-circle arc** facing the camera (concave formation). **Pose & Attire:** All women are sitting on the floor in ‘Seiza’ style (traditional kneeling with legs folded underneath). They are wearing {outer_garment} paired with **fashionable super-short miniskirts**. The look is **attractive and charming** (sexy vibe) but strictly maintains **modesty and safety** (no inappropriate exposure, no upskirt angles). **Action:** They are playfully lifting the bottom hem of their tops with both hands to reveal **clean, solid white T-shirts underneath** (fully clothed layers, no skin exposure on torso). **Text:** On each white T-shirt is a SINGLE, large, bold, handwritten character. Reading from LEFT to RIGHT, the characters spell out \”{text_content}\”. **Critical Diversity:** Each woman must have a COMPLETELY UNIQUE face, distinct hairstyle, and individual expression. Do not generate identical faces or clones. **Environment:** {environment_setting}, with warm atmospheric lighting. 8k resolution. {aspect_ratio}” } |
