this first v1 prototype is trained on fluffyrock.
A current bias is the less simple the background is the less the character will be focused on. ill probably change this since i dont really prefer it.
half of the dataset is survivor and the other scugs dont have enough data bring them out with any decent quality.its easy to invoke survivor and the model biases towards white bodies but its okayish at generating other solid color sluggies as well.
im going to expand the dataset for the other scugs and train it on an anime model so that way more people can use it. ill also more than likely be refining the tags.
dont expect quality outputs for the other unique slugcats for the time being. the goal for the next iteration is to represent them conceptually and to get good outputs for them.
