模型库 /
LoRA /
Hackclaw (麦晓雯) — Delta Force
触发词
hackclaw
文件清单 线路:原始(huggingface.co) · 切换
下载后放进对应的 ComfyUI 目录。models/ 指 ComfyUI 的模型根目录。
说明
Hackclaw (麦晓雯) — Delta Force
作者:00123489abcd968
Hackclaw (麦小雯) — Delta Force Operator LoRA for MOSS-Image
A character LoRA of Hackclaw (麦小雯 / Mai Xiaowen) from Delta Force (《三角洲行动》), trained for MOSS-Image. Trained on a curated mix of in-game realistic CGI renders and cosplay photography using detailed natural-language captions, this LoRA accurately reproduces her signature tactical appearance and flexibly adapts between sleek game CGI and photorealistic photography. ---
Trigger Word & Key Visual Prompts
- Trigger Word: hackclaw
- Hair & Face: short silver-gray bob haircut with bangs, side braid, black tactical face mask covering her nose and mouth, dark focused eyes
- Outfit & Gear: black-and-white patterned scarf, cropped blue and white camouflage jacket with rolled-up sleeves, olive-green tactical vest with multiple pockets, exposed toned midriff, black fingerless gloves, green utility belt with black thigh holster, tight black ripped leather pants, black lace-up combat boots ---
Style Switching Tips
Because the dataset combines game CGI and real-life cosplay references, you can easily steer the visual style using the opening sentence of your prompt:
- Game CGI / Cinematic Style: Start with hackclaw, This is a highly detailed, realistic digital CGI artwork depicting...
- Photorealistic / Cosplay Style: Start with hackclaw, Photograph of a young East Asian woman... ---
Recommended Settings (MOSS-Image)
- Recommended Base Model: MOSS-Image-SFT (works on Base as well; SFT gives cleaner composition and lighting)
- LoRA Weight: 0.85 – 1.0
- Sampler / Scheduler: Euler / sgm_uniform
- Flow Shift: 5.0
- Steps: 50 (or match your Turbo/distilled LoRA step count)
- CFG Scale: 3.0 (keep low for long natural-language prompts)
- Resolution: Multiples of 16 — recommended 896×1344 or 1024×1536 (2:3 portrait)
- Negative Prompt: nsfw, worst quality, bad anatomy, watermark, jpeg artifact
Prompting Note: MOSS-Image uses a Qwen3.5 text encoder. Write clear, descriptive natural-language sentences instead of comma-separated tag spam or CLIP weighting syntax like (word:1.3).
中文简介
本模型为《三角洲行动》干员 麦小雯(Hackclaw) 的 MOSS-Image 角色 LoRA。训练集融合了游戏实机 CGI 与写实 Cosplay 照片,并采用完整自然语言标注:
- 触发词hackclaw
- 推荐底模MOSS-Image-SFT(权重 0.85 ~ 1.0)
- 推荐参数Euler / sgm_uniformShift 5.0Steps 50CFG 3.0
- 风格切换:提示词开头写 This is a realistic digital CGI artwork... 偏向游戏实机/立绘质感;改写为 Photograph of a young East Asian woman... 可自然切换到写实摄影风格。
- 编写提示:MOSS 底模请直接使用流畅的英文自然语言长句描述画面,避免使用 (xxx:1.3) 等旧版权重语法。 If you enjoy this model, feel free to share your generations in the gallery below!
