LoRA is generated entirely from anime images.
-----------------------------------
Note:
I used a wrong activation token TakahashiKK instead of TakahashiM in the sample images. Ignore it. The tag is not even necessary to activate the LoRA.
Check samples for prompts.
-----------------------------------
Leave a like, comment, review but first of all, ENJOY ;)
Description
FAQ
Comments (4)
i really wanna ask you something. Do you use same training method for all LORAs?For instance, your Danmachi LORAs give anime artstyle even at 0.8 weight. But for this anime at 0.8 they only resemble the character features and doesnt really apply anime style. Not only that but i also found out that even at 1.2 weight LORA still functions similarly without changing composition much. Is it different training methods or you just train similar steps and seasonals that do not have enough data end up a bit underbaked (completely speculation on my part)
and yes i did use the correct activation words and tested to make sure lora loads
I changed my settings around when i was making DanMachi... don't remember precisely
Takahashi is probably underbaked due to the low dataset img count and I didn't want to burn her with more repeats than necessary.
There probably won't be any new loras from me, anyways.
I found a different hobby that's actually useful IRL.
I'll probably only make some new cuties from the next season, but not the whole anime sets. No time for that.
@LittleJelly ah. shame your stuff is usually pretty high quality. Have fun with your new hobby. Also if it's not too much to ask can you give me your training settings you used for the Danmachi set? i wanna get into Pony training too but couldn't make any i was satisfied with