metadata
license: cc0-1.0
tags:
- stable-diffusion
- text-to-image
-Specializing in Japanese girls
-Merged photo-real model mixed 5000+ Twitter images.
for Stable Diffusion Webui Automatic1111
type: .safetensors(ckpt)
CFG Scale: middle-low
Negative prompts are rarely needed.
example.
low quality, worst quality, bad anatomy, bad proportions
Recommended Sampler
UniPC, Dpm++ (2M/SDE) Karras, DDIM
Steps: 10～24
Recommended VAE
vae-ft-mse-840000-ema-pruned
Zipang XL version
El Zipang LL
-For LCM & LoRA
Download
### *** important ***
-When using LCM, you should
<lora:pytorch_lora_weights:0.2> (Download)
sampler: DPM++ 2S a Karras (not LCM sampler)
step:6
CFG Scale:3
-When using Myxx series Lora, you should
example:<lora:myjc_v3n:0.5:0.1>
-no Lora
-myxx Lora
-refer to pnginfo
El Zipang K
-Suitable for LoRA
example
-refer to pnginfo
El Zipang
-Mixed 5000+images
- Photo-Real
photo realistic:+++++
portrait:++++
pose:+++
hand:++
example
realistic, photorealistic, detailed, beautiful, RAW photo, film grain, (natural lighting :1.2), (wrinkled skin:1.1),
japanese, 1girl, happy, short black hair, brown eyes, bangs, camisole, close-up, facing viewer, photo background
Negative prompt: (worst quality, low quality, normal quality:1.8), painting, drawing
Steps: 28, Sampler: UniPC, CFG scale: 3.5, Seed: 2272109148, Size: 512x768, Model hash: b7cbea183c, Denoising strength: 0.18, Hires upscale: 2, Hires upscaler: SwinIR_4x
-Recommended prompt: (photoreal tips)
jsx realistic, photorealistic, detailed, beautiful, RAW photo, film grain, (natural lighting, perfect lighting:1.5), wrinkled skin, day, Negative prompt: (worst quality, low quality, normal quality:1.8), painting, drawing, greyscale CFG scale: 3 Hires: 2-4x
Anything4.5(base trained) + uncanny valley(my other model:Any3.0+EZ test series) Brock Weight Merged
Optional (LoRA)
JP Face_v2: Download
768x768 dim512
JP Face: Download
512x512 dim128
-trigger: japanese
cute, realistic, photorealistic, detailed, beautiful, RAW photo, film grain, wrinkled skin, (natural lighting:1.4)
japanese, 1girl, solo, black hair, head tilt, brick wall, looking at viewer, lips, long brown hair, smile, bangs, upper body, turtleneck, long sleeves, white ribbed sweater, brown eyes <lora:jpface:1>
Negative prompt: (worst quality:1.5), bad anatomy, extra digit, greyscale
Steps: 34, Sampler: DPM++ 2M SDE Karras, CFG scale: 4, Seed: 1789550705, Size: 768x512, Model hash: b7cbea183c, Denoising strength: 0.3, Clip skip: 2, Token merging ratio: 0.5, Token merging ratio hr: 0.5, Hires upscale: 2, Hires steps: 30, Hires upscaler: R-ESRGAN 4x+ Anime6B, Lora hashes: jpface: bc0b0c346d61, Version: v1.3.0
How to use:
<lora:jpface:1>
OLD TestVersion
-Specializing in Japanese girls Test Model -Merged photo-real model mixed 1000+ Twitter images.
El Zipang_test1.4
-Mixed 2000+images
-The composition and the human body are even less fragile. Maybe...
test_no_zipang.safetensors + El Zipang_test1.3.safetensors -> Simple Merge!!!
- test no Zipang Anything3.0(base trained) + EZ test1.0 Brock Weight Merged
El Zipang_test1.3
Download
-Just changed the block-weighted merge ratio.
-Better hands and better poses than 1.0. Maybe...
thanks, supermerger!!!
http://github.com/hako-mikan/sd-webui-supermerger/
- EZ test1.0 + Anything3.0 Merged
El Zipang_test1.0(test01)
- Photo-Real
photo realistic:+++++
portrait:+++++
pose:+
hand:+
example
japanese, 1girl, solo, looking at viewer, long hair, couch, black hair, smile, realistic, bangs, long sleeves, black eyes, brown eyes, upper body, sitting, fingernails, lips, school uniform, closed mouth
Negative prompt: low contrast, lowers, normal quality, low quality, worst quality, bad anatomy, bad proportions
Steps: 10, Sampler: DPM++ SDE Karras, CFG scale: 5, Seed: 185236819, Size: 512x768, Model hash: c785dff0