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 s**t 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 s**t content. This issue will be fixed in version 2.