This is an Anima version of my 44 characters from Bungo Stray Dogs LoRA.
It seems to do multi-character images better than the previous versions, but it's still not perfect.
I will probably do even more Anima versions of my past LoRAs in the future.
List of character tags:
nakajima atsushi
dazai osamu
kunikida doppo
edogawa ranpo
yosano akiko
miyazawa kenji
tanizaki junichirou
tanizaki naomi
fukuzawa yukichi
izumi kyouka
akutagawa ryuunosuke
higuchi ichiyou
akutagawa gin
tachihara michizou
mori ougai
elise bsd
nakahara chuuya
kajii motojirou
oda sakunosuke
sakaguchi ango
ozaki kouyou
yumeno kyuusaku
francis fitzgerald
louisa alcott
john steinbeck
howard lovecraft
nathaniel hawthorne
margaret mitchell
mark twain
lucy montgomery
edgar poe
fyodor dostoyevsky
oguri mushitarou
nikolai gogol
fukuchi ouchi
okura teruko
jouno saigiku
suehiro tecchou
sigma bsd
bram stoker
shibusawa tatsuhiko
koda aya
hirotsu ryuurou
agatha christie
A small note/disclaimer: although most of the characters share names with real people, this model was not trained on and does not generate any real people.
Description
Seems to work well, based on my testing
See the model description for trigger words (character tags)
FAQ
Comments (11)
Hi can you do some Characters or justa 1 I need it pls
Can you be more specific? 🙂
@me111 can you do ithinkilikeyou Characters this is a gay furry Comics + free online 😅 can you do it i need them if you can't do all of them just do the bunny guy use the images from season 3
@codira5950666 I'm not familiar with it, but I'll look into it and see what I can do 🙂
@me111 https://ithinkilikeyou.net here+ do it with anima 3 try it with the bunny guy his name is sky
@me111 are you going to do it or not
@Iwsnsiwis282 Sorry, I have been busy. I will make it soon
@me111 oh ok 🤞🙏
Can you share your training setup? I've been trying to train a multi-character lora for anima for a while and it never turns out nearly as good as training on noob. A lot of characters just don't seem to train at all for me.
I'm not an expert by any means so I'm not sure if I can help, but I made sure each character had a roughly equal amount of images in the dataset, and trained for 22k steps on citronlegacy's colab lora trainer
Thanks! I found most of the training details from the lora metadata. Still not quite sure what's going wrong on my side, but I'll try some adjustments.