Experimental merge, some bugs expected. Selecting models according to my personal taste and which in my opinion give good results in terms of generation, v0.1 anime style with pseudo-realism, v0.3 shifted more to semi-realism. In future versions the configurations maybe be more optimal.

Positive prompt：

masterpiece, amazing quality, best quality, ultra-detailed

Positive prompt realism:

masterpiece, amazing quality, best quality, ultra-detailed, realistic, photorealistic

This prompt tags affect generation, for v0.1 more and for v0.3 less:

anime style, toon style, retro style

Negative prompt：

lowres, bad quality, worst quality, bad anatomy, sketch, jpeg artifacts, ugly, poorly drawn, blurry, censored, artist name, signature, watermark

In some cases, you can add to a negative prompt:

wet, wet skin, blush

Recommended settings:

Steps: 25-35

CFG scale: 4-7

Sampler: Euler a / DPM++ 2M

Scheduler: Karras

Model Recommended Resolution: 768x1344, 832x1216, 768x1280, 704x1408, 1024x1536

Used Great Resources:

Also added my Lora in merge process. Merged locally using SD.Next.
Version 0.3 uses updated checkpoints and merged with Strange Thing J.UN.D v0.1toon