Used 29 images of Billie Eilish, tried a new training method including 10 training epochs, a batch size of 2, a 32/32 network configuration, and (10,15) repetitions, without using regularized images. Training was based on sd 1.5 version with a strength setting of 0.8, which should work for most models. Used WD14 captions, completed training in less than 25 minutes. The file size is also smaller. If you prefer this model over the original larger version, please let me know.