Model trained on the British actress Emilia Clarke. I created a full checkpoint because embeddings were just not doing it. They add no new data to a CKPT so the results were average at best. Using 1.5 with new images vastly improves the results. Use the trigger words EmiliaClarke woman to get best results. If you use Emilia Clarke (with the space) you get the 1.5 version of Emilia mixed in, which usually ruins the image but on occasion can help. |