ToB企服应用市场:ToB评测及商务社交产业平台

标题: Stable Diffusion - 常用的负向提示 Embeddings 剖析与 坐姿 (Sitting) 提 [打印本页]

作者: 乌市泽哥    时间: 2024-8-17 18:15
标题: Stable Diffusion - 常用的负向提示 Embeddings 剖析与 坐姿 (Sitting) 提
欢迎关注我的CSDN:https://spike.blog.csdn.net/
本文地址:https://spike.blog.csdn.net/article/details/132145248
  负向 Embeddings 是用于提高 StableDiffusion 生成图像质量的技术,可以避免生成一些不符合预期的图像特征,比如画崩、变形、多余的部位等。原理是将一些不希望出现的图像特征转换为向量表示,然后作为负面提示,输入到 StableDiffusion 算法中,使得算法在优化过程中,只管远离这些向量所代表的图像分布。
测试模型:墨幽人造人_v1030

测试组合:

坐姿的提示词示例:
  1. (masterpiece, best quality:1.2),highly detailed,extremely detailed,real photo,
  2. fullbody,1girl,solo,asian,looking at viewer,(body facing viewer:1.2),(relax sitting),knees separation,
  3. red lips,brown long hair,
  4. collared shirt and dress shirt,long sleeves,(knees length dress:1.1),
  5. (wrap hip very thick pantyhose:1.1),color high heels,
  6. nice figure,good anatomy,good proportions,nice pose,(2shoes,2legs:1.2),(perfect legs:1.1),nice hand,
  7. outdoors,buildings,photorealistic,realistic,<lora:more_details:0.4>,
  8. <lora:yuzuv10:0.5>,<lora:sit_cross_leg_v2:0.6>,<lora:control_skin_exposure:-1.0>,
  9. Negative prompt: (ng_deepnegative_v1_75t:1.3),(negative_hand),(badhandv4),(negative_feet_v2:0.6),
  10. body sideways,buttocks,
  11. extra limbs,malformed limbs,floating limbs,disconnected limbs,missing arm,missing leg,extra arms,extra legs,mutated legs,long neck,
  12. bad anatomy,bad proportions,disfigured,long leg,
  13. worst quality,bad quality,jpeg artifacts,lowres,normal quality,low quality,
  14. EasyNegative,
  15. Steps: 30, Sampler: DPM++ 2M SDE Karras, CFG scale: 5, Seed: 2451060841, Size: 512x768, Model hash: e4a30e4607, Model: 麦橘写实_MajicMIX_Realistic_v6, Denoising strength: 0.35, Clip skip: 2, ADetailer model: face_yolov8n.pt, ADetailer prompt: "asian girl, make up, beautiful face,", ADetailer confidence: 0.3, ADetailer dilate/erode: 4, ADetailer mask blur: 4, ADetailer denoising strength: 0.4, ADetailer inpaint only masked: True, ADetailer inpaint padding: 0, ADetailer ControlNet model: control_v11p_sd15_inpaint [ebff9138], ADetailer ControlNet module: inpaint_global_harmonious, ADetailer version: 23.7.6, Hires upscale: 2, Hires steps: 5, Hires upscaler: 4x-UltraSharp, Lora hashes: "more_details: 3b8aa1d351ef, yuzuv10: b1464588227a, sit_cross_leg_v2: cb80e9bce437, control_skin_exposure: 58bbb7a04626", TI hashes: "ng_deepnegative_v1_75t: 54e7e4826d53, negative_hand: 73b524a2da12, badhandv4: 5e40d722fc3d, negative_feet_v2: df90b1ff666d, EasyNegative: 66a7279a88dd", ControlNet 0: "preprocessor: inpaint_global_harmonious, model: control_v11p_sd15_inpaint [ebff9138], weight: 1.0, starting/ending: (0.0, 1.0), resize mode: ResizeMode.INNER_FIT, pixel perfect: True, control mode: ControlMode.BALANCED, preprocessor params: (-1, -1, -1)", Version: v1.5.1
复制代码
  坐姿提示词很难完全乐成,只管包管大部分可用,即可。
  墨幽真人 模型的效果:

1. BadDream + UnrealisticDream + FastNegativeV2 (DreamShaper)

DreamShaper 模型的常见提示词 Embedding 组合,注意 FastNegativeV2 已经包罗 negative_hand 的内容,即:
  1. BadDream, (UnrealisticDream:1.2), FastNegativeV2,
复制代码
  实测,提示词组合对于 DreamShaper 模型的效果较好,但是墨幽模型的效果不并理想。
  参考地址:

FastNegativeV2 的介绍:

图像示例:

BadDream + UnrealisticDream 的介绍:

图像示例:

2. ng_deepnegative_v1_75t

   测试之后,非常好用的肢体扭曲的提示词,兼容各种模型。
  参考地址:Deep Negative V1.x,更新时间,2023.6.20,ng_deepnegative_v1_75t 的介绍:

图像示例:

3. bad-picture-chill-75v

参考地址:bad-picture negative embedding for ChilloutMix,更新时间,2023.5.8,bad-picture-chill-75v 的介绍:

示例图像:

4. EasyNegative

参考地址:EasyNegative,更新时间,2023.2.10,EasyNegative 的介绍:

示例图像:
![Img5

5. negative_hand + badhandv4

经典的手部修复,参考地址:

negative_hand 的介绍:

示例图像:

badhandv4 的介绍:

示例图像:

6. negative_feet_v2

   测试效果不稳定,需要与 (negative_feet_v2:0.6),(deformed limbs and arms and legs:1.2) 同时使用。
  参考地址:Negfeet : improve feet quality (The Third post),发布时间,2023.6.18,negative_feet_v2 的介绍:


免责声明:如果侵犯了您的权益,请联系站长,我们会及时删除侵权内容,谢谢合作!更多信息从访问主页:qidao123.com:ToB企服之家,中国第一个企服评测及商务社交产业平台。




欢迎光临 ToB企服应用市场:ToB评测及商务社交产业平台 (https://dis.qidao123.com/) Powered by Discuz! X3.4