AgainMix v2.1是一款基于Stable Diffusion 1.5开发的亚洲真人风格生成模型,专门优化和改进了图像生成技术,旨在生成更加真实和细腻的亚洲人面孔和身体特征。通过深度学习和大量训练数据,模型能够精确捕捉亚洲人种的独特外貌,包括肤色、五官比例和面部表情等细节,广泛应用于数字艺术创作、虚拟现实、动画制作等领域,帮助用户轻松生成栩栩如生、富有个性化的亚洲真人形象。
v2.2修改了模型
Realistic models接近真实感的模型,
Merge models 这是个合并的模型
初步的图片将会生成图片的录屏可以参考
图片一般需要有 TI Embedding 文件才能生成同样的图片
作品参照
girl, suit, smirk, realistic skin, skin pores, upper body, film grain, realistic, photorealistic,
Negative prompt: (watermark:1.2), (text:1.2), (logo:1.2), (3d render:1.2), drawing, painting, crayon
Steps: 20, CFG scale: 8, Sampler: DPM++ 2M, Seed: 965654902, VAE: vae-ft-mse-840000-ema-pruned.safetensors, NGMS: 2, Size: 512x768, Model: AgainMix_v2.2, Version: v1.10.0, Model hash: e84c77e38e, Hires steps: 10, Hires upscale: 2, Schedule type: Karras, Hires upscaler: 4x-UltraSharp, ADetailer model: face_yolov8m.pt, ADetailer version: 24.11.1, Denoising strength: 0.45, ADetailer mask blur: 4, Token merging ratio: 0.2, ADetailer confidence: 0.3, ADetailer dilate erode: 4, Token merging ratio hr: 0.2, ADetailer inpaint padding: 32, ADetailer denoising strength: 0.4, ADetailer inpaint only masked: True, Clip skip: 2
girl, suit, waist up, film grain, realistic, photorealistic,
Negative prompt: (watermark:1.2), (text:1.2), (logo:1.2), (3d render:1.2), drawing, painting, crayon
Steps: 20, CFG scale: 8, Sampler: DPM++ 2M, Seed: 3962417508, VAE: vae-ft-mse-840000-ema-pruned.safetensors, NGMS: 2, Size: 512x768, Model: AgainMix_v2.2, Version: v1.10.0, Model hash: e84c77e38e, Hires steps: 10, Hires upscale: 2, Schedule type: Karras, Hires upscaler: 4x-UltraSharp, ADetailer model: face_yolov8m.pt, ADetailer version: 24.11.1, Denoising strength: 0.45, ADetailer mask blur: 4, Token merging ratio: 0.2, ADetailer confidence: 0.3, ADetailer dilate erode: 4, Token merging ratio hr: 0.2, ADetailer inpaint padding: 32, ADetailer denoising strength: 0.4, ADetailer inpaint only masked: True, Clip skip: 2
boy,upper body,
Negative prompt: (worst quality:1.5),(low quality:1.5),(normal quality:1.5),(poor quality:1.5),nsfw,lowres,polar lowres,monochrome,grayscale,low res,drawing,ugly,anime,illustration,
Steps: 30, CFG scale: 10, Sampler: DPM++ 2M, Seed: 1695743237, VAE: vae-ft-mse-840000-ema-pruned.safetensors, NGMS: 2, Size: 512x768, Model: AgainMix_v2.2, Version: v1.10.0, Model hash: e84c77e38e, Hires steps: 10, Hires upscale: 2, Schedule type: Automatic, Hires upscaler: 4x-UltraSharp, ADetailer model: face_yolov8m.pt, ADetailer version: 24.11.1, Denoising strength: 0.3, ADetailer mask blur: 4, Hires schedule type: Karras, Token merging ratio: 0.2, ADetailer confidence: 0.3, ADetailer dilate erode: 4, Token merging ratio hr: 0.2, ADetailer inpaint padding: 32, ADetailer denoising strength: 0.4, ADetailer inpaint only masked: True, Clip skip: 2
作者: | tin18688783 |
最后更新: | 2024年4月16日 |
文件类型: | CHECKPOINT MERGE |
基本模型: | SD 1.5 |
模型版本: | v2.1 |
使用介绍
v2.1修改了模型
Realistic models接近真实感的模型,
Merge models 这是个合并的模型
初步的图片将会生成图片的录屏可以参考
图片一般需要有 TI Embedding 文件才能生成同样的图片
作品参照
Stable Diffusion提示词
1girl,
Negative prompt: (worst quality:1.5),(low quality:1.5),(normal quality:1.5),lowres,watermark,ng_deepnegative_v1_75t,EasyNegative,badhandv4,
Steps: 31, VAE: vae-ft-mse-840000-ema-pruned.safetensors, NGMS: 2, Size: 512x768, Seed: 3152640639, Model: AgainMix_v2.1, Version: v1.7.0, Sampler: DPM++ 2M Karras, CFG scale: 7, Clip skip: 2, Model hash: db05c1ecf8, Hires steps: 10, Hires upscale: 2, Hires upscaler: 8x_NMKD-Superscale_150000_G, Denoising strength: 0.4, Token merging ratio: 0.2, Token merging ratio hr: 0.2
作者: | tin18688783 |
最后更新: | 2024年2月27日 |
文件类型: | CHECKPOINT MERGE |
基本模型: | SD 1.5 |
tag标签: | 真人 亚洲 可爱 摄影 照片 写真 女性 帅哥 美女 |
模型版本: | v1.9 |
使用介绍
v1.0 is a little brighter, v1.0版本颜色更亮一点,
v1.1 is a little darker, v1.1 颜色更暗一点
v1.2 is a little darker, v1.2版本颜色更暗一点,
v1.2 better generate males, v1.2更好的生成男性,
v1.2 has more skin texture (closer to the real person skin),
v1.2版本皮肤纹理更多一点(更接近真人效果),
v1.3 version enhances skin texture, v1.3版本加强皮肤纹理
v1.4 Looks younger. v1.4樣子更年輕一點
v1.5 looks update v1.5樣子更新
v1.6 remodeled the model and added skin texture, breasts +
v1.6重新了模型也增加了皮肤纹理,r房加大
(v1.6建议使用R-ESRGAN General WDN 4xV3 放大算法,速度更快)
v1.7 remodeled the model and breasts +
v1.7重新了模型和r房加大
v1.8 remodeled the model
v1.8重新了模型
(https://github.com/xinntao/Real-ESRGAN/releases)
Realistic models 接近真實感的模型,
Merge models 這是個合併的模型
最早的圖片會有生成圖片的錄屏可以參考
圖片一般需要有 T.I. Embedding 文件才能生成一樣的圖片
作品参照
Stable Diffusion提示词
1girl,
Negative prompt: (worst quality:2),(low quality:2),(normal quality:2),(poor quality:2),lowres,polar lowres,((monochrome)),((grayscale)),low res,drawing,ugly,anime,illustration,
Steps: 20, VAE: vae-ft-mse-840000-ema-pruned.safetensors, NGMS: 2, Size: 512x768, Seed: 3152640639, Model: AgainMix_v1.9, Version: v1.7.0, Sampler: DPM++ 2M Karras, VAE hash: 235745af8d, CFG scale: 7, Clip skip: 2, Model hash: aebc27691d, Hires steps: 10, Hires upscale: 2, Hires upscaler: 8x_NMKD-Superscale_150000_G, Denoising strength: 0.4, Token merging ratio: 0.2, Token merging ratio hr: 0.2
作者: | tin18688783 |
最后更新: | 2023年2月16日 |
文件类型: | CHECKPOINT MERGE |
基本模型: | SD 1.5 |
tag标签: | 真人 亚洲 可爱 摄影 照片 写真 女性 男性 女孩 |
模型版本: | v1.8 |
使用介绍
v1.0 is a little brighter, v1.0版本颜色更亮一点,
v1.1 is a little darker, v1.1 颜色更暗一点
v1.2 is a little darker, v1.2版本颜色更暗一点,
v1.2 better generate males, v1.2更好的生成男性,
v1.2版本皮肤纹理更多一点(更接近真人效果),
v1.3 version enhances skin texture, v1.3版本加强皮肤纹理
v1.4 Looks younger. v1.4樣子更年輕一點
v1.5 looks update v1.5樣子更新
v1.6重新了模型也增加了皮肤纹理,r房加大
(v1.6建议使用R-ESRGAN General WDN 4xV3 放大算法,速度更快)
v1.7重新了模型和r房加大
v1.8重新了模型
(https://github.com/xinntao/Real-ESRGAN/releases)
Realistic models 接近真實感的模型,
Merge models 這是個合併的模型
最早的圖片會有生成圖片的錄屏可以參考
圖片一般需要有 T.I. Embedding 文件才能生成一樣的圖片
作品参照
Stable Diffusion提示词
1girl,
Negative prompt: (worst quality:2),(low quality:2),(normal quality:2),(poor quality:2),lowres,polar lowres,((monochrome)),((grayscale)),low res,drawing,ugly,anime,illustration,
Steps: 20, VAE: vae-ft-mse-840000-ema-pruned.safetensors, NGMS: 2, Size: 512x768, Seed: 3152640639, Model: AgainMix_v1.8, Version: v1.7.0, Sampler: DPM++ 2M Karras, VAE hash: 735e4c3a44, CFG scale: 7, Clip skip: 2, Model hash: de8781b101, Hires steps: 10, Hires upscale: 2, Hires upscaler: 8x_NMKD-Superscale_150000_G, Denoising strength: 0.4, Token merging ratio: 0.2, Token merging ratio hr: 0.2