Type | |
Stats | 1,449 |
Reviews | (381) |
Published | Apr 13, 2023 |
Base Model | |
Hash | AutoV2 C46CCEF059 |
This is a merged model that excels in 2.5D anime style.
It was created to test AutoMBW as well.
Guide
The image above was generated using the following settings
VAE should be kl-f8-anime2.vae
I generated the images using the following settings
Steps: 30 ~ 40,
Sampler: DPM++ SDE Karras,
CFG scale: 15,
Size: 512x704,
Clip skip: 2,
I usually use CFG Scale Fix when generating images.
Dynamic thresholding enabled: True,
Mimic scale: 7,
Threshold percentile: 95,
Mimic mode: Half Cosine Up,
Mimic scale minimum: 0,
CFG mode: Half Cosine Up,
CFG scale minimum: 3.5
This is not necessarily the best setting.
You may change it according to your preference.
Tips
Combined with a very nice LoRA, the result may be even better
There are many other great LoRAs out there
Recommended negative embedding is also listed here