Trained on 83 images created solely by prompts and bodysliders, 18 vectors, learning rate 0.004, 15 steps pr image/embedding and 1800 steps in total. Responds well to all manner of checkpoints.
