The style of ponkotsu has changed so i train a new model. The model is trained by pics he made since 2023.The 1.2 version has more image repeats but less specific tags then V1.1 when training. I wonder which one will work better.warning: It's my mistake. During the training process, images containing st content were directly imported, and the labeling software failed to identify this content and associate it with common prompt words like "anal insertion". As a result, inputting those prompt words may occasionally generate st content. This issue will be fixed in version 2.
