🐰 BUNNY’s SFW Wildcards
LoRA-Friendly English Version | Total ~16,582 Prompts | ~5.65M Tokens
✨ Phase 1 Release / 第一部分正式发布 T
o get this massive dataset into your hands faster, I’m releasing it in three parts. Part 1 (approx. 5,000 prompts) is available now! The remaining two parts will be released continuously after testing is complete.
为了让大家尽快用上，庞大的数据集将分三部分陆续上线。本次率先发布第一部分（约 5,000 条）！ 另外两部分在进一步测试完毕后会陆续上架。
All prompts in Part 1 have been fully cleaned and translated into English via Codex individually.
第一部分的所有提示词均已通过 Codex 完成逐条独立清洗与完整英文翻译。
🧹 LoRA-Friendly Processing / LoRA 友好处理
Prompts were processed individually to strip out redundant character tags while preserving exact scene context: 逐条精细化处理，彻底移除干扰项，完美适配人物 LoRA：
❌ Removed (已清理): Facial features, hairstyle/color, race, and age descriptions (无五官/发型/发色/种族/年龄描写，无空行及中文残留).
✅ Preserved (已保留): Actions & poses, clothing, emotions, environments, composition, perspective, lighting & atmosphere (保留动作、服饰、情绪、场景、镜头构图及光影氛围).
📊 Pack Statistics / 数据统计
Total Prompts (总条数): 16,582
Average (平均长度): 255.4 words (~340 tokens)
Median / P95 (中位数/P95): 263 words (~350 tokens) / 322 words (~430 tokens)
Max Length (最长单条): 502 words (~670 tokens)
Total Volume (总计): ~5.65 Million Tokens
🧪 Best For / 推荐用途
🎲 Random portrait generation & large-scale batching (随机人像生成与大规模跑图)
👤 Character & Style LoRA testing (人物/画风 LoRA 测试)
📷 Scene and composition exploration (场景与构图探索)
🤖 Models with strong natural-language understanding (适配对长自然语言理解力强的模型)
(Note: Prompts are dense and detailed. Shorten manually if using models sensitive to context length.) (注：提示词信息密度较高，对 Prompt 长度敏感的模型建议手动精简)
🔮 Next Up / 后续发布计划
I am currently testing the next two specialized parts, which will be uploaded soon: 另外两个特定需求的版本正在测试中，随后发布：
🔞 Mild NSFW Clean Version (轻度 NSFW 清洁版): Tastefully filtered and stripped of character tag conflicts.
(保持克制与精细清洗，剥离干扰，适配特定 LoRA 出图)
🏞️ Pure Scene & Background Version (纯背景/场景版): Stripped of all character descriptions. Gives you complete control over character counts (single/multiple) and overall character design while maintaining highly detailed, atmospheric background environments.
(彻底剥离所有人物描述，方便完全自由地控制画面人数与整体人物设计，同时保留高细节与氛围感的大背景。)
GPT-Converted English Version | 380 Entries | ~940K Tokens
🔗 My LoRA Series / 我的 LoRA 系列
Since I had already converted such a large amount for personal testing, I decided to share it publicly as well. 🐰
既然已经为了自用测试转换了这么大的体量，我也决定顺便公开分享。
📌 Original Post / 原帖
🔗 Original SFW15000 Wildcard Post
The original Chinese wildcard collection comes from the post above.
原始中文版 Wildcard 来源于上方帖子。
🤖 About This Version / 关于此版本
The prompts were converted from Chinese into fluent English using GPT, with the goal of preserving as much of the original detail, structure, scene description, composition, lighting, camera perspective, materials, and visual atmosphere as possible.
这些提示词使用 GPT 从中文转换为流畅英文，目标是尽可能保留原文中的细节、结构、场景描述、构图、光影、镜头视角、材质与整体画面氛围。
This is not a simple keyword extraction or tag conversion.
这并不是简单的关键词提取或 TAG 转换。
The original long-form natural-language structure is retained as much as possible, including:
✨ Character appearance
👗 Outfits
🕺 Poses and actions
😊 Expressions
🏙️ Environments
📐 Composition
📷 Camera perspective
💡 Lighting
🧵 Materials and textures
🌫️ Depth of field
🎨 Overall visual atmosphere
英文版尽可能保留原始长篇自然语言 Prompt 的信息密度与描述结构。
⚠️ ANIMA Warning / ANIMA 注意事项
❌ ANIMA is not recommended.
❌ 不推荐用于 ANIMA。
In my testing, these prompts are simply too long and information-dense for ANIMA.
根据我的测试，这类提示词对于 ANIMA 来说过长、信息密度过高。
Very long prompts can easily cause:
🌫️ Blurry generations
⚠️ Reduced image quality
🌀 Unstable composition
📉 Overall visual degradation
超长提示词很容易导致：
🌫️ 画面变糊
⚠️ 质量下降
🌀 构图不稳定
📉 整体视觉效果劣化
So while the wildcard file can technically be loaded into an ANIMA workflow, I do not recommend using the full unmodified prompts directly.
因此，虽然 Wildcard 文件 technically 可以加载进 ANIMA 工作流，但我不建议直接使用完整的原始长提示词。
🔞 Upcoming NSFW Modified Version / 即将发布的 NSFW 修改版
The previously mentioned DeepSeek NSFW version is now finished and released.
之前在 SFW 版本中提到的 DeepSeek NSFW 版本现在已经完成并正式发布。
NSFW Wildcards | Light-to-Medium | 11 TXT Files | ~1.6M Tokens Total
🐰 BUNNY’s Heavy NSFW Wildcards| LoRA Friendly
✅ Content / 内容说明
This release contains SFW content only.
当前版本仅包含 SFW 内容。
Suitable for:
🎲 Random prompt generation
👤 Portrait generation
🤳 Selfie generation
🧪 Workflow testing
🐰 BUNNY’s LoRA testing
💡 Prompt inspiration
📚 Dataset inspiration
🤖 Natural-language image model testing
适合用于随机提示词生成、人像生成、自拍生成、工作流测试、LoRA 测试、Prompt 灵感、数据集灵感以及自然语言图像模型测试。
🐰 BUNNY’s Series
More LoRAs, sliders, concepts, testing tools, and wildcard resources will be added over time.
后续会继续发布更多 LoRA、滑块、概念模型、测试工具与 Wildcard 资源。