There's an existing SD1.5 model I used some generated images from, plus direct images. Trained on a total or 71 images. Strength of 1 usually works, but there is more flexibility with lower values.