Agnetha was trained on 70 AI prompted images. The embedding is pretty good, but for some reason holds little resemblance to the source.