SD1.5 : recommended setting is 0.6
Description
trained with updated dataset
FAQ
Comments (8)
What is that model you're using for this and do you remember the merge recipe for it? Just curious because the style is pretty nice and aesthetic
Also I'm really surprised to see this character get a LORA model. Cudos to you sir
firstly, thankyou and yes i remember the merge :
model (A) notRealAnimeStyle_v18 : (Not-Real-Anime-Style)
model (B) childrensStories_v1CustomA : (Childrens Stories)
Interpolation Method : weighted Sum
Multiplier (M) : 0.3
that the merge model i used for the preview.
for the training model of V3 , i use AnyLoRA_noVae_fp16
Very well done Lora. It creates a very consistent look for the character. Include the word "child" in your prompt to get her to look close to her proper age of 12. Many of the checkpoints I tried want to make her look too old. This is not the fault of the Lora, just a strong tendency of many models toward making girls look in their late teens. I included two examples of this in my image post below.
I appreciate you taking time to write a comment and I am glad you liked my model. Although I tried to train multiple Model, some details are not generated consistently, like the red hair string. Maybe I should remove it from the dataset Images?
When I have enough good quality images, I will try to train another model that will be more consistent across other checkpoints.
Civitai does not allow "child" as a prompt onsite tho
@Murdochella
you can maybe try "Young" , i don't really use the site generator.
@Jabberwocky207 Infact the "two examples of this in my image post below" you said are gone, I presume due to CivitaAI censorsship...
@settima_ai
I think you wanted to reply to Jabberwocky207, not me.