Embedding trained on 30 images of actress Ana de Armas. I know, there are already several TIs, LORAs, hypernetworks, what have you, of this actress already (including one of mine). This is just to replace my older hypernetwork if anyone is interested. It was trained on mostly the same dataset, but I did remove a couple of pictures that had an outfit that negatively affected the output from the HN. If anyone is curious, it wa the banded metal dress from the UK premiere of The Gray Man. (Not the LA premiere, that dress was fine). In the older HN, it would show up as vertical stripes on all kinds of outfits as well as a shimmering rainbow/oil slick coloring effect on clothes, even when it didn't have the metal banding.
Description
base version.
FAQ
Comments (4)
THX，I REALLY LIKE IT
You always have the best TI‘s any chance of a tutorial ?
I can't remember which off the top of my head, but I gave a pretty detailed answer to a comment on one of the other embeddings a few weeks back, maybe 2. If not a comment on this one, check the Barbara Palvin one. I'll look real quick, but it was basically following a tutorial on yourube/reddit with a variable learning rate, them following up with testing various step-iterations to find one that hadn't over trained then resuming training from that point with a lower learning rate to refine the embedding. I'm finding that I might have tobretrain this one and Daisy, though, because while they worked fine in RinaMix2, they are not as good in RinaMix3, which I hope to have up this weekend.
Correction, it was a question asked on the Avril Lavigne embedding.
edit: link to comment below. I will need to retrain them, because even though all training is done in SD1.5, my previous TIs were TESTED in RinaMix2, and refined for that model in particular, because at the end of the day, I made these for me, and that's the general use model I was using. Since I'm using a new merge I'll now need to retrain at least 3; Ana, Daisy and Robo Faces, though that one can still be made to work, it just takes a lot of fiddling with the prompt.
https://civitai.com/models/10053/avril-lavigne-embedding?commentId=31128&modal=commentThread
Anyway, good luck and have fun!
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.