
权重:0.5-0.8,推荐0.7,权重越高,3D感越强
推荐使用:高分辨率修复 (Hires. fix)
【强烈推荐使用模型】:OnlyAnime |《唯》· 炫彩动漫 https://civitai.com/models/105955/onlyanime-or
vae: kl-f8-anime2
【Trigger Words】
1.正面提示词:qingCB, 1girl, yellow eyes, brown hair, ponytail,
正面可选提示词:bodysuit, pantyhose, hair ornament, Metal arm ring, gloves, armor, boots
2.负面提示词:3D,3D rendering
负面可选提示词:weapon, holding weapon, gun, handgun, poor light
3.负面embaddings: UnrealisticDream, EasyNegative,
【一个参考预设,您可以在此之上进行修改】
1.正面提示词:qingCB, 1girl, solo, looking at viewer, yellow eyes, sparkling eyes, brown hair, ponytail, hair ornament, bodysuit, pantyhose, breasts, full body, masterpiece, best quality, (colorful),(finely detailed beautiful eyes and detailed face),cinematic lighting, extremely detailed CG unity 8k wallpaper, <lora:qingCB-v1:0.7>
2.负面提示词:EasyNegative, UnrealisticDream,3D, 3D rendering, poor light, weapon, holding weapon, handgun,
3.其它参数:DPM++SED Karras, 30 steps ,512x768, 启用Hires.fix, R-ESRGAN 4X+Anime6B, 放大1.25倍,重绘幅度0.35
【目前存在的问题与建议】
1.在创作NSFW内容时,请根据实际情况降低权重至0.5-0.6,同时使用更多的NSFW相关提示词,增加NSFW相关提示词的权重,并将bodysuit等衣服类Trigger Words 加入负面提示词,
2.当您感觉图片细节不足、模糊时,请适当增加迭代步数,或调整采样器及vae,
3.绘制远景全身图片时,请适当增加出图尺寸与迭代步数.
4.该Lora的权重越低,越应该提高人物特征类【Trigger Words】的权重,例如:yellow eyes, brown hair, ponytail
5.强烈推荐使用负面embaddings:UnrealisticDream, EasyNegative
6.可能存在的模型泛用性问题
————————————————————————————————————————————————————————
【Trigger Words】
1. Positive prompt words: qingCB, 1girl, yellow eyes, brown hair, ponytail,
Optional prompts on the front: bodysuit, pantyhose, hair ornament, Metal arm ring, gloves, armor, boots
2. Negative prompt words: 3D, 3D rendering
Negative optional prompt words: weapon, holding weapon, gun, handgun, poor light
3. Negative embaddings: UnrealisticDream, EasyNegative,
【A reference preset that you can modify on top of】
1. Positive prompt words: qingCB, 1girl, solo, looking at viewer, yellow eyes, sparkling eyes, brown hair, ponytail, hair ornament, bodysuit, pantyhose, breasts, full body, masterpiece, best quality, (colorful), (finely detailed beautiful eyes and detailed face), cinematic lighting, extremely detailed CG unity 8k wallpaper, <lora:qingCB-v1:0.7>
2. Negative prompt words: EasyNegative, UnrealisticDream, 3D, 3D rendering, poor light, weapon, holding weapon, handgun,
3. Other parameters: DPM++SED Karras, 30 steps, 512x768, enable Hires.fix, R-ESRGAN 4X+Anime6B, zoom in 1.25 times, redrawing range 0.35
【current problems and suggestions】
1. When creating NSFW content, please reduce the weight to 0.5-0.6 according to the actual situation, and at the same time use more NSFW-related prompt words, increase the weight of NSFW-related prompt words, and add bodysuit and other clothing trigger words to negative prompt words.
2. When you feel that the details of the picture are insufficient or blurred, please increase the STEPS, or adjust the sampler and vae.
3. When drawing a full-body picture of a distant view, please increase the size of the picture and the number of iteration steps appropriately.
4. The lower the weight of the Lora, the more the weight of the character feature [Trigger Words] should be increased, for example: yellow eyes, brown hair, ponytail
5. It is strongly recommended to use negative embaddings: UnrealisticDream, EasyNegative
6. Possible model generalization problems
描述:
训练词语: qingCB, 1girl, yellow eyes, brown hair, ponytail, bodysuit, pantyhose, hair ornament,
名称: qingCB-v1.safetensors
大小 (KB): 73840
类型: Model
Pickle 扫描结果: Success
Pickle 扫描信息: No Pickle imports
病毒扫描结果: Success