新功能:真实感的整体提升,我终于发现了导致 aDetailer错误和图像错误的原因:是括号()
因此,如果您想避免面部或图像生成错误,请避免在正面和负面提示上使用(或尽可能少地使用它们)
特别注意:我建议将 ADetailer 的默认设置更改为 0.5 去噪强度和分辨率 768×768 像素。
非常重要:如果您的提示中存在冲突(例如,[full body photo] 而另一部分有 [face photo],或者模型无法理解的内容,这可能会导致图像生成失败。因此,请密切注意您的提示
- 我发现的最佳配置
- VAE: 一个特殊的 VAE 已经包含在内,你不需要它 – (感谢nuaion)
- Upscaler: 4x-Ultrasharp或4X NMKD Superscale
- Clip skip: 2
- 宽度: 768
- 身高: 1344
- CFG:5.5
- Steps:30
- 采样最喜欢的方法: DPM++ 2M SDE (更真实,但速度慢)或 DPM++ 2M (不太真实,但速度更快)
- 调度员: Karras
- 一般提示示例:
- 积极提示: score_9, score_8_up, score_7_up, (Type of Shoot), (Subject), the description, (background), more details
- 示例: score_9, score_8_up, score_7_up, Raw Photo, Cowboy Shot, yorha no. 2 type b, short white hair, black dress, hairband, black gloves, clothing cutout, cleavage cutout, puffy sleeves, long sleeves, black hairband, feather-trimmed sleeves, juliet sleeves, mole under mouth, holding katana, black thighhighs, turtleneck, sunset, rubble in the background, depth of field, dynamic angle, photo realistic:1.4, realistic skin:1.4, fashion photography, sharp, analog film grain, hyperdetailed:1.15
- 负面提示:score_6, score_5, score_4, cartoon, painting, illustration, 3D rendering, CG, ugly, old, fat, wide hips, curvy, topless, nude, naked, nsfw, child
触发词: | Raw Photo Realistic skin:1.4 photo realistic:1.4 |
使用提示: | CLIP SKIP: 2 |
作品参照
score_9, score_8_up, score_7_up, Portrait, girl, egyptian queen, cute, seductive, innocent, light smile:0.3, plump lips, slender body, black hair, bob cut, perfect make-up, intricate jewelry, golden headdress, captured in a walls adorned with hieroglyphics, depth of field, dynamic angle, photo realistic:1.4, realistic skin:1.4, fashion photography, sharp, analog film grain
Negative prompt: score_6, score_5, score_4, cartoon, painting, illustration, 3D rendering, CG, ugly, old, fat, topless, nude, naked, nsfw, child, runny makeup
Steps: 35, CFG scale: 5, Sampler: DPM++ 2M SDE, Seed: 1408358398, RNG: NV, Size: 768x1344, Model: AniMerge-Pony-XL-VAE-v3.0.fp16, Version: v1.10.1, Model hash: 0c7712870d, Schedule type: KL Optimal, ADetailer model: face_yolov8n.pt, ADetailer version: 24.8.0, ADetailer mask blur: 4, ADetailer confidence: 0.3, ADetailer dilate erode: 4, ADetailer inpaint width: 1024, ADetailer inpaint height: 1024, ADetailer inpaint padding: 32, ADetailer denoising strength: 0.45, ADetailer inpaint only masked: True, ADetailer use inpaint width height: True, Clip skip: 2
score_9, score_8_up, score_7_up, Raw Photo, Full Body Shot, Ayanami Rei, cute, seductive, innocent, light smile:0.3, plump lips, slender body, red eyes, short light blue hair, floating hair, white plugsuit, sitting, sunset, rubble in the background, depth of field, dynamic angle, photo realistic:1.4, realistic skin:1.4, fashion photography, sharp, analog film grain, hyperdetailed:1.15
Negative prompt: score_6, score_5, score_4, cartoon, painting, illustration, 3D rendering, CG, ugly, old, fat, wide hips, curvy, topless, nude, naked, nsfw, child, ironman, armor
Steps: 35, CFG scale: 5, Sampler: DPM++ 2M SDE, Seed: 3768894768, RNG: NV, Size: 768x1344, Model: AniMerge-Pony-XL-VAE-v3.0.fp16, Version: v1.10.1, Model hash: 0c7712870d, Schedule type: KL Optimal, ADetailer model: face_yolov8n.pt, ADetailer version: 24.8.0, ADetailer mask blur: 4, ADetailer confidence: 0.3, ADetailer dilate erode: 4, ADetailer inpaint width: 1024, ADetailer inpaint height: 1024, ADetailer inpaint padding: 32, ADetailer denoising strength: 0.45, ADetailer inpaint only masked: True, ADetailer use inpaint width height: True, Clip skip: 2
score_9, score_8_up, score_7_up, Portrait, girl, venezuelan, cute, seductive, innocent, light smile, plump lips, slender body, atmospheric scene, solo, midriff, navel, brown hair, outdoors, standing, pink headband, track background, depth of field, dynamic angle, photo realistic:1.4, realistic skin:1.4, fashion photography, sharp, analog film grain
Negative prompt: score_6, score_5, score_4, cartoon, painting, illustration, 3D rendering, CG, ugly, old, fat, topless, nude, naked, nsfw, child
Steps: 35, CFG scale: 5.5, Sampler: DPM++ 2M, Seed: 1408358398, RNG: NV, Size: 768x1344, Model: AniMerge-Pony-XL-VAE-v3.0.fp16, Version: v1.10.1, Model hash: 0c7712870d, Schedule type: KL Optimal, ADetailer model: face_yolov8n.pt, ADetailer version: 24.8.0, ADetailer mask blur: 4, ADetailer confidence: 0.3, ADetailer dilate erode: 4, ADetailer inpaint width: 1024, ADetailer inpaint height: 1024, ADetailer inpaint padding: 32, ADetailer denoising strength: 0.45, ADetailer inpaint only masked: True, ADetailer use inpaint width height: True, Clip skip: 2
使用介绍
关于此版本
重要提示:与 1.0 版本相比,正面和负面提示已发生改变:
整体改善
- 提高真实感
- 改良型男性
- 改进的机器人
特别说明:我建议将 ADetailer 的默认设置更改为 0.5 去噪强度和 768×768 像素。
非常重要:如果您的提示中存在冲突(例如,[全身照片] 而另一部分有 [脸部照片],或者模型无法理解的内容,这可能会导致图像生成失败。因此,请密切注意您的提示
VAE: 一个特殊的 VAE 已经包含在内,你不需要它 – (感谢nuaion)
Clip skip: 2
CFG: 5~5.5
Steps: 30
Sampling favorite method: DPM++ 2M
Scheduler: KL Optimal or Karras
一般提示示例:积极提示:score_9, score_8_up, score_7_up, (Type of shot), (Subject), [the description], (background), more details, (depth of field), (dynamic angle), (fine textures details), (realistic:1.4), (realistic skin:1.4), (sharp), (artistic photo), (analog film grain)
负面提示:score_6, score_5, score_4, cartoon, painting, illustration, 3D rendering, CG, Unreal engine, fake, plastic, doll, ugly, old, fat, grayscale
举例: score_9, score_8_up, score_7_up, (Cowboy shoot photo), (Morrigan Aensland), cute, seductive, innocent, green hair, floating hair, luscious lips, black belt, (sitting on a white cloud), crossed legs, bat wings, pink tights with bat prints, starry sky, detailed, (depth of field), (dynamic angle), (fine textures details), (realistic:1.4), (realistic skin:1.4), (sharp), (artistic photo), (analog film grain)
触发词: | (RAW photo) (realistic skin 1.4) (realistic:1.4) (analog film grain) cute, seductive, innocent |
使用提示: | CLIP SKIP: 2 |
A1111我的设置:
我使用以下设置运行我的家用电脑A1111:
- 设置 COMMANDLINE_ARGS= –xformers –skip-torch-cuda-test –no-half-vae
(如果您的 VRam 较低,请尝试添加–medvram -sdxl或–lowvram,这可以帮助您,但它会减慢图像创建速度)
如果您无法安装 xFormers(请参阅下文)请使用我的Google Colab 设置:
- 设置 COMMANDLINE_ARGS= –disable-model-loading-ram-optimization –opt-sdp-no-mem-attention –no-half-vae
(如果您的 VRam 较低,请尝试添加–medvram -sdxl或–lowvram ,它们可以帮助您,但它会减慢图像创建速度)
我的 A1111 版本:版本:v1.9.3 • python:3.10.11 • torch:2.1.2+cu121 • xformers:0.0.23.post1 • gradio:3.41.2 •
如果您想像我的家用电脑 一样激活 xformers 优化(如何安装 xFormers):
- 在 A1111 中单击“设置选项卡”
- 在左侧栏中,点击“优化”
- 在:“交叉注意优化”中选择:“ xformers ”
- 按“应用设置”
- 重启稳定扩散
如果你无法安装 xFormers,请使用 SDP-ATTENTION,例如我的 Google Colab:
- 在 A1111 中单击“设置选项卡”
- 在左侧栏中,点击“优化”
- 在:“交叉注意力优化”中选择:“ sdp-no-mem – 无记忆高效注意力的缩放点积”
- 按“应用设置”
- 重启稳定扩散
如何模拟 nvidia GPU 请按照以下步骤操作:
- 在 A1111 中单击“设置选项卡”
- 在左侧栏中,点击“显示所有页面”
- 搜索“随机数生成器源”
- 选择语音:“ NV ”
- 按“应用设置”
- 重启稳定扩散
如果您使用我的模型,请为您的 A1111安装ADetailer扩展。
导航到Stable Diffusion 中的“扩展”选项卡。
- 转到“从 URL 安装”子部分。
- 在“扩展的 git 存储库的 URL”下输入此链接: https: //github.com/Bing-su/adetailer
- 点击“安装”按钮安装扩展
- 重启稳定扩散
如何安装 Euler Max Sampler:
- 在 A1111 中单击“扩展选项卡”
- 点击“从 URL 安装”
- 在“扩展的 git 存储库的 URL”下输入此链接:https://github.com/licyk/advanced_euler_sampler_extension
- 安装后点击:“已安装”选项卡
- 点击“应用并退出”
- 重启稳定扩散
- 现在,在采样器列表的末尾,您有了新的采样器。
- VAE: 一个特殊的 VAE 已经包含在内,你不需要它 – (感谢nuaion)
- Clip skip: 2(仍在测试)
- Upscaler: 4x-Ultrasharp或4X NMKD Superscale
- 宽度: 768-(可能更多,但我没有测试过)
- 高度: 1280-(可能更多,但我没有测试过)
- CFG Scale: 5~7(可能也是从 3 开始,但我还没有测试过)
- 步数:20~30
- 我发现的最好的
- CFG:6
- 步骤:30
- 采样最喜欢的方法: DPM++ 2M
- Scheduler:SGM Uniform
- 一般提示示例
- 积极提示: score_9, score_8_up, score_7_up, (RAW photo), (type of shot)(Subject), [the description], (background), more details.
- 示例:(RAW photo), (upper body photo), (Native American Woman), 20yo, slim, skinny, adorned with traditional regalia and feathers, standing, (Canyon background), (intricate details), (dynamic angle), (cinematic lights), seductive, innocent, (realistic skin:1.4), slim, skinny, detailed Native American tribal dress, sharp, (analog film grain)
- 负面提示: score_6, score_5, score_4, cartoon, painting, illustration, 3D rendering, CG, Unreal engine, fake, plastic, doll, ugly, old, fat, grayscale
HiRes.Fix 设置:
我不使用 Hi.Res Fix,因为:
1)在我的电脑上不工作
2) 我的模型不需要它。使用 txt2image、aderailer 和资源选项卡中建议的升级器。
如果您仍想使用它,这是MarkWar发给我的设置 。
Hires upscale: 1.5
Hires steps: 20~30
Hires upscaler: R-ESRGAN 4x + Anime6B,
Denoising strength: 0.4
Adetailer: face_yolov8n
该模型可免费供个人 使用,并可免费供个人合并( * )。
如需商业使用,请务必联系我(Ko-fi)或通过电子邮件联系:samuele[dot]bonzio[at]gmail[dot]com
作品参照
score_9, score_8_up, score_7_up, (Portrait photo), (Jinx), cute, seductive, innocent, cloud tattoos on arm, outdoor, intense blue long hair, bokeh, neon light passing through hair, award winning fashion photography, (city background), (depth of field), (dynamic angle), (fine textures details), (realistic:1.4), (realistic skin:1.4), (sharp), (artistic photo), (analog film grain)
Negative prompt: score_6, score_5, score_4, cartoon, painting, illustration, 3D rendering, CG, ugly, old, fat, topless, nude, naked, nsfw, source_furry, source_pony, source_cartoon
Steps: 30, CFG scale: 5.5, Sampler: DPM++ 2M, Seed: 4233988065, RNG: NV, Size: 768x1344, Model: AniMerge-Pony-XL-VAE-v2.0.fp16, Version: v1.10.1, Model hash: 671c7b7f5a, Schedule type: Karras, ADetailer model: face_yolov8n.pt, ADetailer version: 24.8.0, ADetailer mask blur: 4, ADetailer confidence: 0.3, ADetailer dilate erode: 4, ADetailer inpaint width: 768, ADetailer inpaint height: 768, ADetailer inpaint padding: 32, ADetailer denoising strength: 0.45, ADetailer inpaint only masked: True, ADetailer use inpaint width height: True, Clip skip: 2
score_9, score_8_up, score_7_up, (Portrait photo), Cammy White, seductive, innocent, blonde hair, twin braids, red beret, blue eyes, green leotard, sleeveless, camouflage, (light smile:0.3), rays of sunshine through fog, (swampy forest background), (depth of field), (dynamic angle), (fine textures details), (realistic:1.4), (realistic skin:1.4), (sharp), (artistic photo), (analog film grain)
Negative prompt: score_6, score_5, score_4, cartoon, painting, illustration, 3D rendering, CG, ugly, old, fat, topless, nude, naked, nsfw, source_furry, source_pony, source_cartoon
Steps: 30, CFG scale: 5.5, Sampler: DPM++ 2M, Seed: 910106442, RNG: NV, Size: 768x1344, Model: AniMerge-Pony-XL-VAE-v2.0.fp16, Version: v1.10.1, Model hash: 671c7b7f5a, Schedule type: Karras, ADetailer model: face_yolov8n.pt, ADetailer version: 24.8.0, ADetailer mask blur: 4, ADetailer confidence: 0.3, ADetailer dilate erode: 4, ADetailer inpaint width: 768, ADetailer inpaint height: 768, ADetailer inpaint padding: 32, ADetailer denoising strength: 0.45, ADetailer inpaint only masked: True, ADetailer use inpaint width height: True, Clip skip: 2
score_9, score_8_up, score_7_up, (Full body photo), (Ayanami Rei), cute, seductive, innocent, red eyes, short light blue hair, sitting on green grass, luscious lips, white Evangelion plugsuit, (lake background), (depth of field), (dynamic angle), (fine textures details), (realistic:1.4), (realistic skin:1.4), (sharp), (artistic photo), (analog film grain)
Negative prompt: score_6, score_5, score_4, cartoon, painting, illustration, 3D rendering, CG, ugly, old, fat, topless, nude, naked, nsfw, source_furry, source_pony, source_cartoon, ironman, armor
Steps: 30, CFG scale: 5.5, Sampler: DPM++ 2M, Seed: 4233988064, RNG: NV, Size: 768x1344, Model: AniMerge-Pony-XL-VAE-v2.0.fp16, Version: v1.10.1, Model hash: 671c7b7f5a, Schedule type: Karras, ADetailer model: face_yolov8n.pt, ADetailer version: 24.8.0, ADetailer mask blur: 4, ADetailer confidence: 0.3, ADetailer dilate erode: 4, ADetailer inpaint width: 768, ADetailer inpaint height: 768, ADetailer inpaint padding: 32, ADetailer denoising strength: 0.45, ADetailer inpaint only masked: True, ADetailer use inpaint width height: True, Clip skip: 2
使用介绍
关于此版本
我必须承认,做这个 AniMerge 比想象中要难,为了找到最正确的设置,我花了 4 天的时间尝试并创建了 2000 多张图像。
最后,我不得不说,它让我很满意,我已经达到了自己设定的半现实主义水平。缺点是,在提示方面它有点“困难”。
话虽如此,我还是把设置留给你了。
触发词: | (RAW photo) (hyperrealistic photo) seductive innocent (realistc skin:1.4) |
使用提示: | CLIP SKIP: 2 |
A1111我的设置:
我使用以下设置运行我的家用电脑A1111:
- 设置 COMMANDLINE_ARGS= –xformers –skip-torch-cuda-test –no-half-vae
(如果您的 VRam 较低,请尝试添加–medvram -sdxl或–lowvram,这可以帮助您,但它会减慢图像创建速度)
如果您无法安装 xFormers(请参阅下文)请使用我的Google Colab 设置:
- 设置 COMMANDLINE_ARGS= –disable-model-loading-ram-optimization –opt-sdp-no-mem-attention –no-half-vae
(如果您的 VRam 较低,请尝试添加–medvram -sdxl或–lowvram ,它们可以帮助您,但它会减慢图像创建速度)
我的 A1111 版本:版本:v1.9.3 • python:3.10.11 • torch:2.1.2+cu121 • xformers:0.0.23.post1 • gradio:3.41.2 •
如果您想像我的家用电脑 一样激活 xformers 优化(如何安装 xFormers):
- 在 A1111 中单击“设置选项卡”
- 在左侧栏中,点击“优化”
- 在:“交叉注意优化”中选择:“ xformers ”
- 按“应用设置”
- 重启稳定扩散
如果你无法安装 xFormers,请使用 SDP-ATTENTION,例如我的 Google Colab:
- 在 A1111 中单击“设置选项卡”
- 在左侧栏中,点击“优化”
- 在:“交叉注意力优化”中选择:“ sdp-no-mem – 无记忆高效注意力的缩放点积”
- 按“应用设置”
- 重启稳定扩散
如何模拟 nvidia GPU 请按照以下步骤操作:
- 在 A1111 中单击“设置选项卡”
- 在左侧栏中,点击“显示所有页面”
- 搜索“随机数生成器源”
- 选择语音:“ NV ”
- 按“应用设置”
- 重启稳定扩散
如果您使用我的模型,请为您的 A1111安装ADetailer扩展。
导航到Stable Diffusion 中的“扩展”选项卡。
- 转到“从 URL 安装”子部分。
- 在“扩展的 git 存储库的 URL”下输入此链接: https: //github.com/Bing-su/adetailer
- 点击“安装”按钮安装扩展
- 重启稳定扩散
如何安装 Euler Max Sampler:
- 在 A1111 中单击“扩展选项卡”
- 点击“从 URL 安装”
- 在“扩展的 git 存储库的 URL”下输入此链接:https://github.com/licyk/advanced_euler_sampler_extension
- 安装后点击:“已安装”选项卡
- 点击“应用并退出”
- 重启稳定扩散
- 现在,在采样器列表的末尾,您有了新的采样器。
- VAE: 一个特殊的 VAE 已经包含在内,你不需要它 – (感谢nuaion)
- Clip skip: 2(仍在测试)
- Upscaler: 4x-Ultrasharp或4X NMKD Superscale
- 宽度: 768-(可能更多,但我没有测试过)
- 高度: 1280-(可能更多,但我没有测试过)
- CFG Scale: 5~7(可能也是从 3 开始,但我还没有测试过)
- 步数:20~30
- 我发现的最好的
- CFG:6
- 步骤:30
- 采样最喜欢的方法: DPM++ 2M
- Scheduler:SGM Uniform
- 一般提示示例
- 积极提示: score_9, score_8_up, score_7_up, (RAW photo), (type of shot)(Subject), [the description], (background), more details.
- 示例:(RAW photo), (upper body photo), (Native American Woman), 20yo, slim, skinny, adorned with traditional regalia and feathers, standing, (Canyon background), (intricate details), (dynamic angle), (cinematic lights), seductive, innocent, (realistic skin:1.4), slim, skinny, detailed Native American tribal dress, sharp, (analog film grain)
- 负面提示: score_6, score_5, score_4, cartoon, painting, illustration, 3D rendering, CG, Unreal engine, fake, plastic, doll, ugly, old, fat, grayscale
HiRes.Fix 设置:
我不使用 Hi.Res Fix,因为:
1)在我的电脑上不工作
2) 我的模型不需要它。使用 txt2image、aderailer 和资源选项卡中建议的升级器。
如果您仍想使用它,这是MarkWar发给我的设置 。
Hires upscale: 1.5
Hires steps: 20~30
Hires upscaler: R-ESRGAN 4x + Anime6B,
Denoising strength: 0.4
Adetailer: face_yolov8n
该模型可免费供个人 使用,并可免费供个人合并( * )。
如需商业使用,请务必联系我(Ko-fi)或通过电子邮件联系:samuele[dot]bonzio[at]gmail[dot]com
作品参照
score_9, score_8_up, score_7_up, (RAW photo), (upper body photo), seductive, (Alita), (Battle Angel Alita), mechanical arms, full cyborg body parts, glowing, sitting, bobbed hair, floating hair, narrow hips, slim, skinny, adult, flirting, (detailed cybernetic body), (light smile:0.3), full lips, glossy, softlight passing through hair, (futuristic city background), (dynamic angle), (cinematic lights), (intricate details), (dynamic angle), (cinematic lights), dark mood, dystopia, (intricate details), (dynamic angle), (cinematic lights), seductive, innocent, (realistic skin:1.4), sharp, (analog film grain)
Negative prompt: score_6, score_5, score_4, cartoon, painting, illustration, 3D rendering, CG, Unreal engine, ugly, old, fat, topless, nude, naked, nsfw, grayscale
Steps: 30, CFG scale: 6, Sampler: DPM++ 2M, Seed: 2953175234, RNG: NV, Size: 768x1280, Model: AniMerge-Pony-XL-VAE-v1.0.fp16, Version: v1.9.4, Mask blur: 4, FP8 weight: Enable, Model hash: d410f21a4d, Inpaint area: Only masked, Schedule type: Karras, ADetailer model: face_yolov8n.pt, ADetailer version: 24.1.2, Denoising strength: 0.5, ADetailer mask blur: 4, Masked area padding: 32, ADetailer confidence: 0.3, ADetailer dilate erode: 4, ADetailer inpaint width: 768, ADetailer inpaint height: 768, ADetailer inpaint padding: 32, Cache FP16 weight for LoRA: True, ADetailer denoising strength: 0.5, ADetailer inpaint only masked: True, ADetailer use inpaint width height: True, Clip skip: 2
score_9, score_8_up, score_7_up, (RAW photo), woman in (Elsa of Arendelle) detailed cosplay, cowboy shot, dark cape, dark dress, fur trim, single braid, snowflake pattern, (light smile:0.3), full lips, glossy, snow, (intricate details), (dynamic angle), (cinematic lights), seductive, innocent, (realistic skin:1.4), sharp, (analog film grain)
Negative prompt: score_6, score_5, score_4, cartoon, painting, illustration, 3D rendering, CG, Unreal engine, ugly, old, fat, topless, nude, naked, nsfw, grayscale
Steps: 30, CFG scale: 6, Sampler: DPM++ 2M, Seed: 3866362322, RNG: NV, Size: 768x1280, Model: AniMerge-Pony-XL-VAE-v1.0.fp16, Version: v1.9.3, Mask blur: 4, Model hash: d410f21a4d, Inpaint area: Only masked, Schedule type: Karras, ADetailer model: face_yolov8n.pt, ADetailer version: 24.1.2, Denoising strength: 0.5, ADetailer mask blur: 4, Masked area padding: 32, ADetailer confidence: 0.3, ADetailer dilate erode: 4, ADetailer inpaint width: 768, ADetailer inpaint height: 768, ADetailer inpaint padding: 32, ADetailer denoising strength: 0.5, ADetailer inpaint only masked: True, ADetailer use inpaint width height: True, Clip skip: 2
score_9, score_8_up, score_7_up, (RAW photo), skinny fashion cute italian model, (Rei Ayanami), white evangelion bodysuit, flirting, light smile:0.3, full lips, glossy, head tilt, detailed face, (sunset, lake background), (intricate details), (dynamic angle), (cinematic lights), seductive, innocent, (realistic skin:1.4), sharp, (analog film grain)
Negative prompt: score_6, score_5, score_4, cartoon, painting, illustration, 3D rendering, CG, Unreal engine, ugly, old, fat, topless, nude, naked, nsfw, grayscale
Steps: 30, CFG scale: 6, Sampler: DPM++ 2M, Seed: 829238465, RNG: NV, Size: 768x1280, Model: AniMerge-Pony-XL-VAE-v1.0.fp16, Version: v1.9.3, Mask blur: 4, Model hash: d410f21a4d, Inpaint area: Only masked, Schedule type: Karras, ADetailer model: face_yolov8n.pt, ADetailer version: 24.1.2, Denoising strength: 0.5, ADetailer mask blur: 4, Masked area padding: 32, ADetailer confidence: 0.3, ADetailer dilate erode: 4, ADetailer inpaint width: 768, ADetailer inpaint height: 768, ADetailer inpaint padding: 32, ADetailer denoising strength: 0.5, ADetailer inpaint only masked: True, ADetailer use inpaint width height: True, Clip skip: 2