Must read: In V8.3 version, retraining is carried ou* ***ng the Cog natural language TAG labeling training

Extracting part of the images in the large model training set for LOHR training and integrating into the large model

Support for image output (headshots, portraits, half-body photos, full-body photos)

Iterative deployment: (25 steps-30 steps)

Hint word guidance coefficient CFG scale: (7)

Recommended sampling method:DPM++ 2S a

Recommended enlargement algorithm: 4x-UltraSharp (or) 4x_NMKD-Superscale-SP_178000_G

(Suggesting a redraw range of about 4) (High score iterative steps: 25 steps-30 steps)

(Run locally -- enlarging is recommended at 1.5-2x) (Very suitable for running large horizontal wallpaper)