
This model was trained on over 100 images that were meticulously tagged by hand as verbosely as possible on a slow learning rate.
In general this lora works best at strengths around 0.45-0.75 depends on the base model used, and sizes like 1024x512 or 512x1024, depending on if you want a portrait or landscape image. works fine with 512x512 as well but can be more messy and make more mistakes. base models like furry vixens usually works best (anime models tend to be burned for anime characters and produce worse results ironically)
This model has a lot of highly specific trigger tags which allow you to highly customize your output. See the example images for prompt examples, the trigger words are
lolibomber - the general trigger, included on all images as the first tag in training and set to not shuffle. If you are looking to make alt costumes or change base traits like hair color then use only this tag and none of the others. doing that you can easily make her a white hair freckle covered girl without it fighting you too much.
megu_staff_full, megu_staff_mid, megu_staff_top, megu_staff_bottom -
tags related to the staff, broken into sections to be as hyper particular as possible. for example if you want her holding it behind her back then do not include megu_staff_full or megu_staff_mid. If you want her pointing it towards the screen then do not include full, mid or bottom, etc. This will help the model realize where it needs to be. not perfect and still sometimes has issues placing it but usually can get what you what with a few batches
megu_outfit_shirt - relates to her normal dress/skirt outfit can easily be removed or added without the rest of the outfit without issue using this tag
megu_outfit_cape - just her cape, can easily be removed or added without the rest of the outfit without issue using this tag
megu_outfit_hat - the hat, can easily be removed or added without the rest of the outfit without issue using this tag, removed from head sadly usually removes is from the image entirely, i am hoping in updates to make it possible to have it in the image but just on the ground or in her hand. might need a new tag for it
megu_outfit_boots - the usual boots, can easily be removed or added without the rest of the outfit without issue using this tag
megu_outfit_gloves - the usual fingerless gloves, can easily be removed or added without the rest of the outfit without issue using this tag
megu_outfit_eyepatch - this one is not quite as clean as the rest, it sometimes drags the top half of her dress with it, which for full outfit is fine but if you wanted say, nude but eyepatch can sometimes take some cycles to get it. will hopefully be improved in future updates to the model
描述:
initial version
训练词语: megu_staff_full,megu_staff_top,megu_staff_mid,megu_staff_bottom,megu_outfit_shirt,megu_outfit_cape,megu_outfit_hat,megu_outfit_boots,megu_outfit_gloves
名称: megumin-LK-V1.safetensors
大小 (KB): 36991
类型: Model
Pickle 扫描结果: Success
Pickle 扫描信息: No Pickle imports
病毒扫描结果: Success