LoRA training cost Civitai Buzz. If you like my work, please consider tipping me so I can keep creating them. Reacting to my images, posting your images and using Civitai generator with my models can also help.
Description
FAQ
Comments (4)
Hello, I have previously trained the Laura model for an anime character, but the results always contain many errors and are very bad because I did not modify the training parameters of the custom model: AnyLoRA Anime Mix. Is this the reason??
it can be many reasons. how many images in your dataset? how you tag the images? do you train for pony or sdxl or sd 1.5 ? you use civitai for training or other tool? how many epochs?
@andinmaro146 The number of photos is approximately between 5 and 6
I Put tags according to the words in the image description box
I train on sd1.5
Number of eras: 10
Yes, use civitai for training
@afdlljyl453 i use 50-150 images for anime characters and the character don't need to be solo. i use wd 1.4 autotagger for tagging with booru tags. i also remove tags that related to the character like eye or hair color and style and add the character name at the start. i don't use civitai training for sd 1.5 but for pony the default parameters are good so i guess they configure it for sd 1.5 as well.
i use anime screencap for my lora and train with AAM for characters and AnyLora for styles.
i prepare the datasets like this https://civitai.com/models/22530/guide-make-your-own-loras-easy-and-free