The V8 version is different from previous versions, the large-scale models have been retrained, and the latest GPT-4 low-quality labeling and Cog natural language completion details are used. At the same time, retraining on hand micro-adjustments, light sensitivity, and color, integrated into this retrained large model.

Recommended iteration steps:25 steps - 30 steps

Recommended sampling method:DPM++ 2M Karras

The name of the local webui 1.9.0:DPM++ 2M

Recommended enlargement algorithm:R-ESRGAN_4x+ Anime6B

Enable After Detailer when needed:face_yolov8n_v2.pt