I was wondering... is possible to do character LoRAs?
I was wondering... is possible to do character LoRAs?
Fyes
Fof course
Chow would you go about creating a custom character? as in the dataset?
CLet's consider an example where I try different prompts from scratch and then generate one example that I like
Cdo you think some augmentation of the data would hlep me create some LoRA and then try to generate some off that and then I have some checkpoints which will help me generate more and more ?
Cis that nuts 
?
?
ADr going over that video to anime tutorial. How are you able to apply this logic of "concepts" inside onetrainer for example?
Fadd 2 concepts
Fyou can make their repeating accordingly their sizes
Fyou can use trained checkpoint to generate images
Fand train again
Ffrom those generated images
Cthanks man
Cyou're a star... my best results came from your tutorial and I've tried a few
Fjust wait till you try fine tuning
For you tried it?
CNope; I just did a LoRA with flux
Cwith my face
Cis that what you mean by fine tuning or ?
Fok then you will likem it
Ffine tune / dreambooth
Frecording tutorial now
Cniceeeeeeeee
CFyep
Fvideo recording over 2 hours already :d
Ccan't wait 
Fthis is like a full course tutorial haha
M@Dr. Furkan Gözükara time for you to stay away and run more tests.
lycoris training available now for kohya and he also fixed the prior preservation loss
lycoris training available now for kohya and he also fixed the prior preservation loss
Mhe will merge into the branch in a bit
Fi will more tests later i need to publish tutorial first
Frest will be only updating configs
MLycoris already looks quite good with dora on flux, so hopefully you can explore better settings when you do
XI can’t wait for this ahahah + can wait for 4K image to video!
BHello all- I am looking for help with a project relating to flux and creating seamless images. Right now, it is unable to make images that can be tiled into larger designs. If anyone is up for a project, I am willing to pay for its development.
Adr for the style, what happens if you add in like I dont know 5,000 images for the stylization then you do half that number for your character. Is that too much?
Dhttps://huggingface.co/nyanko7/flux-dev-de-distill Unet LR 0.0001 vs 0.00003 the loos curve looks so different, up 80 epochs at 0.0001 it started to overtrain, lets see now...
Dmay be 0.00003 is too low, idk
DIsn't this like, already 7.5 times higher than what the Doc recommended?
Dno
Dthis is one of the Dr. Furkan config files 3e-05=0.00003
Dthat's with the T5 options enable, may be I have to increase it to 0.00005 in this file is at that value
DOhh, I see. You're doing a lora, not a full dreambooth. My bad.
Dyes lora
DI will test a finetune in the future with flux-dev-de-distill but it uses cfg scale 3.5 that is slower, I can try to extract the lora and try it with regular flux-dev at full speed
DYes, makes sense
PTesting Style transfer function in new ui, is being born.
Dand do you recommend masked training over "normal" training?
Dalso as far as I remember your dataset is images without a detailed background anyway, right?
Njust had amazing results training a flux fine tune , thank u