No need for complex prompts, simple prompts can also make high quality images.

Highly recommended for use in combination with the series of YFilter LoRAs(see my homepage).

YFilter_Pure_Color: adjust the overall atmosphere and brightness.

YFilter_Dynamic LDR/HDR filter: adjusting the black and white dynamic range.

YFilter_GazeHorizontal: control/fix eye contact.

YFilter_MakeupLevel: decrease or increase pink makeup.

*tip: using YFliter LoRas only within the extension AfterDetail (to regenerate face), you can adjust the face (panning the eyes, increasing or decreasing the makeup, etc.) without changing anything else in the original image at all.

Introduce:

1.Realistic SDXL checkpoint.

2.No need quality prompts, very friendly for beggingers. Simple prompts can make high quality images.

3.More real (face) skin texture, less AI-like feelings.

Recommend Usage:

Important：

1.Don't use quality prompts, including both positive and negative, like: best quality, worst quality, etc. They may not work as your wish. If you have to use quality prompts, please do a comparison test.

2.Bulit-in VAE.

3.Do not use sd_xl_refiner.

Sampling Method (Sampler):

1.DPM++ 2M Karras: ≥ 30 steps

2.Restart: ≥ 20 steps

Guidance Coefficient (CFG Scale):(about) 8

Resolution:The base resolution of SDXL is 1024*1024. Usually the total number of pixels should be around 1024*1024 for a good result. If you are restricted by hardware conditions, it may be appropriate to lower the resolution a little, but not too much.

(1:1) 1024*1024

(4:5) 896*1120

(3:4) 864*1152

(2:3) 832*1248

(3:5) 768*1280

(9:16) 720*1280

(1:2) 704*1408

High-Resolution Restoration (Hires. fix): 4x-UltraSharp or 8x-NMKD-Superscale, denoise ≤0.2.