--------------------------------------------Divider-----------------------------------------

24.2.09:

Today is New Year's Eve, I wish everyone a happy new year, may good luck come in the Year of the Dragon!

Updating two versions at once. The example images of the two versions are comparative images with the same parameters. If interested, you can check the differences yourself.

It may take a while for the next update, the next update should be a major adjustment for version 1.3.X.

Version 1.2.3 fix mainly because when creating the images before, it was found that some backgrounds or clothing could not be reproduced. For someone like me who almost never uses Lora's pure tag flow, I found it unacceptable, so I made some minor adjustments. The skin color and face shape have been slightly altered, and there are changes to the prompt word effects. And 1.2.3 2.9D is an unexpectedly blended version that incorporates elements of 2.8D, but the proportion of control is relatively small. It's just sitting around on the hard drive, so I'll show it to friends who like this kind of effect.

The files are uploaded first, the example images will be updated when available. For now, the original, non-enlarged images generated locally are displayed. The example images of the two versions are generated with the same parameters.

(It should be noted that the cover image generated locally cannot be completely replicated online with lib. This is due to the difference in GPU CPU seeds between local and online, not a problem with the model itself.)

Local parameters: DPM++2M Karras     Iteration steps: 25    cfg: 15  Resolution: 512X912

Plugins enabled: AD Face Retouching, Dynamic Thresholding (CFG Scale Fix)

No high-resolution repair    One of the example images has cn's openpose enabled