✨ Prompt Guide (提示词指南)
1. 🟢 Core Trigger & Appearance (核心触发与外貌)
Trigger Word:
hack_dfEssential Features (必填特征):
blonde hair,black-framed glassesNote: Glasses are detachable! Remove the tag
black-framed glassesto see him without eyewear.注意:眼镜是可拆卸的!如果不写眼镜的提示词,你可以生成他不戴眼镜的样子。
2. 🛡️ Default Tactical Gear (满配特工服)
Best for combat scenes. The chest rig details (patches, tools, protein bar) are highly responsive. 适合战斗场景。胸挂上的细节(贴章、工具、蛋白棒)响应非常精准。
Prompts: hack_df, 1boy, blonde hair, black-framed glasses, black high-collar shirt, grey short-sleeved jacket, black chest rig, tactical vest, goggles on chest, purple lenses, morale patches, tactical tools, protein bar, pouches
3. 👕 Green Shirt Skin (绿色轻量战术服)
The casual/lightweight alternative outfit. 休闲轻量化的绿色战术半拉链衫造型。
Prompts: hack_df, 1boy, blonde hair, black-framed glasses, quarter-zip green shirt, rolled-up sleeves, tactical gloves, olive green tactical pants
4. 💻 Tech/Hacker Mode (极客模式)
stripping down to the inner layer for workshop scenes. 脱去战术背心,只保留内搭,适合维修或黑客场景。
Prompts: hack_df, 1boy, blonde hair, black-framed glasses, black high-collar shirt, bare arms, holding glowing electronic device
⚙️ Recommended Parameters (推荐生成参数)
LoRA Weight (权重):
0.8 - 1.0Since we separated the outfit tags, keeping the weight high helps with the complex gear details.
由于我们对服装进行了拆分打标,较高的权重有助于保持装备细节的清晰度。
Base Resolution (基础分辨率):
1024 x 1024(Recommended / 推荐)Sampler (采样方法):
DPM++ 2M KarrasorEuler aSteps (迭代步数):
25 - 35CFG Scale (提示词引导系数):
7Negative Prompts (推荐负面词):
grey hair, black hair, (low quality:1.4)Add "grey/black hair" to negative prompt to ensure his hair stays blonde in dark environments.
建议在负面词中加入"grey/black hair",防止在暗光环境下发色变黑。
💬 Feedback (反馈)
If you notice any issues during generation (such as the fox patch not rendering correctly or hands holding devices awkwardly), please let me know or share your images in the comments. I will analyze your showcases to optimize and improve the next version of this model. Let's make it even better together!
如果你在出图时发现任何瑕疵(比如狐狸贴章不够清晰、手持设备姿势奇怪等),请务必带图在评论区告诉我。我会仔细参考大家的返图,以此作为下一个版本 (V2.0) 优化迭代的最重要依据!
