<h1 id="bunny's-nsfw-wildcards">🐰 BUNNY’s NSFW Wildcards</h1><h1 id="update-heavy-nsfw-part-2-or-couple-interaction">UPDATE — Heavy NSFW Part 2 | Couple Interaction</h1>Added 1,663 LoRA-friendly couple interaction wildcards with randomized pose, partner traits, expression, clothing state, and fluid combinations. 新增 1,663 条 LoRA 友好型双人互动 Wildcards，随机组合体位、男性属性、女性表情、服装状态与液体表现。This section is designed as a LoRA-friendly wildcard system, focused on interaction concepts rather than replacing anatomy quality from the base model. 这一部分是 LoRA 友好型 Wildcard 系统，主要负责提供互动概念与组合变化，并不替代底模本身的人体结构能力The allocation pools cover multiple male heights, body types, appearance styles, penis profiles, 14 common interaction positions, 10 female expressions, 12 original-outfit clothing states, and 10 fluid outcomes. 分配池包含多种男性身高、体型、外貌气质与尺寸轮廓，以及 14 种常规互动体位、10 种女性表情、12 种保留原服装款式的穿着状态和 10 种液体结果。This wildcard provides interaction concepts only and does not repair anatomy by itself. 本 Wildcard 只负责提供双人互动概念，并不会自行修复人体结构。For better results, use a strong NSFW base model or supporting LoRA with reliable two-person anatomy, such as PornMaster. 为了获得更稳定的效果，建议搭配双人结构能力较强的 NSFW 底模或辅助 LoRA，例如 PornMaster。In a continuous test of 100 fully randomized generations, around 80 results were usable; most failures came from complex limb interaction, hands, feet, or anatomy errors. 连续随机测试 100 张中，大约 80 张结果可用；失败结果主要来自复杂肢体互动、手脚或局部结构错误。If your base model has weak two-person anatomy, you may still get limb errors, finger or toe issues, or occasional body merge problems. 如果底模本身双人结构较弱，仍有可能出现肢体错误、手脚问题或身体融合等情况。<hr /><h3 id="heavyhardcore-nsfw-reworked-version-or-solo-or-character-lora-friendly">Heavy/Hardcore NSFW Reworked Version | Solo | Character LoRA Friendly</h3><hr /><h3 id="about-this-version">🔞 About This Version / 关于此版本</h3>This is the refined Heavy NSFW version of my wildcard series, specifically optimized for character LoRA generation. The goal is a controlled NSFW fine-tune on the source text, not a full scene rewrite. It preserves the original scene, composition, camera angle, lighting, and pose foundation as much as possible, while pushing the content into a stronger solo NSFW direction (e.g., Nude, Half-nude, Open/Torn clothing, Displaced outfit state). Heavy NSFW, BDSM, and fetish elements may be fused when suitable.这是我 Wildcard 系列中的精修重度 NSFW 版本，专门针对人物 LoRA 出图进行优化。其核心目标是在原文基础上做可控的 NSFW 微调，而非整段推倒重写。它会最大程度保留原始场景、构图、镜头、光影、氛围和基础姿势，同时将内容推向更强的单人 NSFW 方向（如：全裸、半裸、敞开/撕破服装、拉开/错位穿着等）。在合适的情况下融合 Heavy NSFW、BDSM 与 Fetish 风格元素。<h3 id="release-info-and-credits">📦 Release Info &amp; Credits / 规模与致谢</h3><ul><li>📌 Original Post / 原帖地址: <a target="_blank" rel="ugc" href="https://civarchive.com/models/2231696/sfw15000-and-qwen-and-z-image-or-15000-portrait-and-selfie-wildcards-for-qwen-and-z-image">SFW15000 &amp; Qwen Wildcards</a> — Huge thanks to the original author for creating this detailed prompt collection!</li><li>感谢原作者制作了如此庞大且细节丰富的 Prompt 合集！🙏</li><li>📊 Current Size (Part 1): Contains 1,663 different prompts (~1.45M characters / 360K tokens). The remaining two parts are of similar size and will be released after testing.</li><li>📊 当前发布规模 (第一部分): 包含 1,663条 不同提示词（约 145万字符 / 36万 tokens）。剩余两部分体量基本相同，测试完成后陆续发布。</li></ul><hr /><h3 id="core-rules-and-logic">🎨 Core Rules &amp; Logic / 核心规则与逻辑</h3><ul><li>🐰 Character LoRA Friendly / 人物 LoRA 友好设计</li><li>To avoid fighting against your character LoRA, this version intentionally removes all facial features, face shapes, eye/nose/lip details, makeup, skin details, hairstyles, hair colors, hair lengths, and bangs/ponytail descriptions. The wildcard handles the scene, outfit state, camera, and NSFW action, while your LoRA safely controls the face, hair, and character identity.</li><li>为了避免与人物 LoRA 冲突，本版本刻意去除了所有五官、脸型、眼鼻唇、妆容、发型、发色、发长及刘海等细节描写。Wildcard 仅负责场景、姿势、服装状态与 NSFW 动作，让 LoRA 完美掌控角色面部与身份特征。</li></ul>🔥 Content Rules / 内容规则<ul><li>This version is solo only.</li><li>本版本为 纯单人 SOLO。</li></ul><ul><li>The NSFW actions mainly focus on solo scenarios, including self-play, fingers, toys, vibrators, spreading, solo climax, and fluid-related details.</li><li>NSFW 动作主要围绕单人场景，包括自慰、手指、玩具、震动棒、自己扒开、单人高潮和体液细节等。</li><li>Heavy NSFW, BDSM, and fetish-style elements may be lightly fused into the original scene when appropriate.</li><li>在适合的情况下，会在原始场景中轻度融合 Heavy NSFW、BDSM 与 Fetish 风格元素。</li><li>Each line aims to include visible adult exposure and at least one fluid or mess detail, while still preserving the original scene logic as much as possible.</li><li>每条 Prompt 尽量包含明确的成人暴露内容，并加入至少一种液体或凌乱细节，同时尽可能保留原始场景逻辑。</li><li>👗 Clothing State / 服装穿着状态</li><li>the original outfit is usually preserved and shifted into compromised states: Nude, half-nude, torn clothing, open clothing, or displaced outfit states. This preserves the original character setup instead of turning every prompt into the same generic NSFW outfit.</li><li>通常保留原始服装并改变穿着状态（全裸、半裸、撕破、敞开或错位状态），以此保留原角色的设定与场景风格，避免千篇一律。</li></ul><hr /><h3 id="why-solo-only-solo">🧍 Why Solo Only? / 为什么仅限单人 SOLO？</h3>This version is adult solo only (no partners/multi-character interaction). It was processed via my custom custom API translation/rewriting pipeline using DeepSeek Flash. During testing, multi-person interactions often led to messy limb errors or broken contact logic due to prompt complexity. Rewriting it into a solo-only format vastly improves stability and practical usability.本版本为成人单人 SOLO 专用（不含搭档或多人互动）。整个合集基于 DeepSeek Flash 通过我自制的高并发 API 流程完成“中文改写+英文回译”的双阶段处理。测试中发现多人互动提示词极易导致肢体错误或互动逻辑混乱，因此全面改为纯单人模式，以确保最高稳定性和实用性。<hr /><h3 id="character-lora-friendly-lora">👤 Character LoRA Friendly / 人物 LoRA 友好设计</h3>To avoid fighting against your character LoRA's identity, this version intentionally removes all:❌ Facial-features, face-shapes, eye/nose/lip/skin details, makeup, and expressions.❌ Hairstyles, hair-color, hair-length, bangs, ponytails, or braids. The wildcard handles the environment, clothing state, camera, and NSFW action—while your LoRA carries the character identity.为了避免和人物 LoRA 产生严重的权重冲突，本版本刻意去除了所有：❌ 五官/脸型描写、眼睛/鼻子/嘴唇/皮肤细节、妆容及绝大多数表情描写。❌ 发型、发色、发长、刘海、马尾、辫子等一切头发细节描写。Wildcard 仅全权负责场景、姿势、服装状态、镜头、光影、氛围和 NSFW 动作，让 LoRA 完美控制角色本尊的脸与特征。<hr /><h3 id="practical-guide">🧪 Practical Guide / 使用与兼容性指南</h3><ul><li>⚠️ Model Compatibility / 模型兼容性: Best suited for modern large models that support detailed scene descriptions and long natural-language prompting.</li><li>更适合支持详细场景、镜头语言和长自然语言提示词的现代大型图像生成模型.</li></ul><hr /><h2 id="how-it-was-made">⚙️ How It Was Made / 制作方式</h2>This version was created using my own custom API translation / rewriting pipeline.这个版本使用我自制的 API 翻译 / 改写流程制作。The workflow uses:⚙️ High-concurrency API processing 📄 One source segment sent independently each time 🎲 A recipe pool for randomness 🧩 Different pose recipes 👗 Different clothing-state recipes 🔞 Different NSFW category recipes 🇨🇳 Chinese-side rewriting first 🇬🇧 Then translated back into English流程包括：⚙️ 高并发 API 处理 📄 每段文本单独独立发送 🎲 使用配方池保证随机性 🧩 不同姿势配方 👗 不同服装状态配方 🔞 不同 NSFW 类别配方 🇨🇳 先在中文侧完成修改 🇬🇧 再翻译返回英文This helps keep each prompt independent, reduces repeated patterns, and gives the wildcard collection more random variation.这样可以让每条 Prompt 保持独立，减少重复模式，并让整个 Wildcard 合集拥有更多随机变化。<hr /><h3 id="feedback-and-disclaimer">💬 Feedback &amp; Disclaimer / 反馈与说明</h3>In my testing, around 5% of images may still show anatomy/object placement errors, but it is highly stable for batch runs. If you have unique ideas or prompt logic to improve the next parts, feel free to leave a comment or DM me! 🐰🔥根据实际生图测试，约有 5% 的图可能存在肢体或道具位置异常，但对于批量跑图已足够稳定。如果你有更好的 Prompt 改进建议，欢迎留言或私信，有价值的反思会直接优化到后续章节中！玩得开心！🐰🔥

Files with SHA256: 829816911c1457dba120126485cb38edf76f0c51fb973dc5940e1452b55c6b0c