Trained on 82 images created solely by prompts and bodysliders, 15 vectors, learning rate 0.004, 15 steps pr image/embedding and 1650 steps in total. Responds well to all manner of checkpoints.
