AWPortraitWW 1.1, this time continues to improve in image quality performance, the delicacy of direct output effect is comparable to the XL and DIT architecture models. There has been a significant improvement in semantic understanding, race, and facial diversity. Especially in studio and indoor photography, the image quality performance is indistinguishable from real photos. It is very suitable as a base model for text-to-image, image-to-image, and real-series LoRA training. The realism of LoRA trained based on ww1.1 will be significantly better than other 1.5 base models. It is no exaggeration to say that WW1.1 once again raises the ceiling of the 1.5 architecture.

For the best image quality, please follow the following parameter configuration settings.

Sampling recommendation: DPM++ 2M Karras or Restart, Step 30

Resolution: 512*768 (recommended), 576*864

Upscaling algorithm: Lanczos, Redraw magnitude 0.2-0.3

Negative keywords: ng_deepnegative_v1_75t, (badhandv4:1.2), (worst quality:2), (low quality:2), (normal quality:2), low res, bad anatomy, bad hands, ((monochrome)), ((grayscale)), watermark, moles, large breast, big breast

Adetailer:

Facial model face_yolov8n

Close-up situations, you can choose not to enable or enable a redraw magnitude of 0.25-0.3.

UpperBody half-body situations, you can choose not to enable or enable a redraw magnitude of 0.3-0.35.

Fullbody full-body situations, please enable a redraw magnitude of 0.35-0.4.