QQ|Archiver|手机版|小黑屋|AiOMG ( 粤ICP备2022147020号-1 )

GMT+8, 2024-10-16 20:24 , Processed in 0.034771 second(s), 30 queries .

Powered by Discuz! X3.5 Licensed

© 2001-2024 Discuz! Team.

[Checkpoint] FaceBombMix

26
0
0
0
simonlive LV9
正文
发布时间:2024-08-06
模型类型
模型类型: Checkpoint
基础模型: SD 1.5
最新版本: v1_bakedVAE
触发词: girl
下载链接: https://pan.baidu.com/s/1qHG_1ogZHIw0Mpg6YmYotA?pwd=hivo
提取码: hivo
源链接: https://civitai.com/models/7152?modelVersionId=25993
模型介绍:
FaceBomb : Covers from anime to 2.5D style. Suited for general use.
More info : https://huggingface.co/mocker/KaBoom

EDIT 1
If the color is washed out, try applying VAE.
https://huggingface.co/hakurei/waifu-diffusion-v1-4/blob/main/vae/kl-f8-anime2.ckpt

EDIT 2
▣ Positive : (masterpiece, sidelighting, finely detailed beautiful eyes: 1.2), masterpiece*portrait, realistic, 3d face, lustrous skin,
▣ Negative : (worst quality, low quality:1.4), watermark, logo,

If the sample prompt doen't generate the image you desire, try adding upper body to positive prompt. That 2.5D style best applies to portrait of upper body. It's not necessary but may save you from extra *click.
Here's my configuration for duplicating first image. Sorry for inconvenience.
I use xformers so some detail(e.g. necklass) may change every time.
Below is just an example and you can/should change configuration to your liking. It's recommended but not limited.

Model : FaceBombMix-fp16-no-ema.safetensors
SD VAE : kl-f8-anime2.ckpt
Clip skip : 2
Sampling method : DPM++ SDE Karras
Sampling steps : 32 (24 ~ 32, as you like)
Upscaler : R-ESRGAN 4x+ Anime6B
Hires steps : 14 (in my case, 7 ~ 14 weren't really different for R-ESRGAN 4x+ Anime6B)
Denoising strength : 0.5 (I change around 0.45 ~ 0.55)
Upscale by : 2
CFG Scale : 9 (7 ~ 9, as you like)

(Settings > Sampler parameters > Eta noise seed delta) is set to `31337` unsure if it matters though.
(Sorry, I dont know hot to upload higher resolution image to description.)

EDIT 3(Not needed if you use bakedVAE version)
How to apply VAE
Download VAE from EDIT 1 and copy to 'stable-diffusion-webui/models/VAE'.
Start the webui and go to 'Settings > User interface > Quicksettings list' and add 'sd_vae' without quotation mark.
Click 'Apply settings' and 'Reload UI'. They are orange buttons on top.
You should see SD_VAE dropdown menu on top of your main screen. Select 'kl-f8-anime2'.

~~~~~~~~~~~~~~~~~~~~~~~o(* ̄︶ ̄*)o~~~~~~~~~~~~~~~~~~~~~~~

示例提示词和参数:
Prompt: (masterpiece, sidelighting, finely detailed beautiful eyes: 1.2), masterpiece*portrait, realistic, 3d face, glowing eyes, shiny hair, lustrous skin, solo, embarassed

Negative prompt: (worst quality, low quality:1.4), watermark, logo,

Steps: 32, CFG scale: 9, Sampler: DPM++ SDE Karras, Seed: 3624413002, ENSD: 31337, Size: 512x768, Model: __Custom_FaceBombMix-fp16-no-ema, Model hash: aad629159b, Hires steps: 14, Hires upscale: 2, Hires upscaler: R-ESRGAN 4x+ Anime6B, Denoising strength: 0.5, Clip skip: 2

本帖子中包含更多资源

您需要 登录 才可以下载或查看,没有帐号?立即注册 微信登录 手机动态码快速登录

x
回复

使用道具

 
您需要登录后才可以回帖 登录 | 立即注册 微信登录 手机动态码快速登录

本版积分规则

快速回复 返回顶部 返回列表