Kelsey Remige

Done by request. If you have any requests, feel free to let me know.

V1 was accidentally trained without any image captioning but still provided good results. V2 is re-trained with almost identical settings but with image captioning, but actually gives me worse results. Putting both up to give you the choice of whichever works best for you.

Strength of about 0.9 seemed to work best for me.

As with all models of real people, don't be weird.