Maybe it's a bit too much, but can you give me the lora?
desu guen
2024-06-29 14:38:20 +0000 UTC
The original PNG file with info.【https://mega.nz/file/sQBhQATL#JbXdGkaVMYoZMnXvPiEe3yeUUfz3639Xz8LESehHnEc】
NFKK
2024-06-29 02:31:10 +0000 UTC
Removing the background is not always necessary; it is generally used when training characters. If you don't mind background style contamination, this step can be skipped. Additionally, when training the art style, removing the background is usually not required.
NFKK
2024-06-29 02:15:01 +0000 UTC
character CG from the game 'Path to Nowhere' without background。However, the default transparent background of the PNG file needs to be replaced with a solid white background.This step can be quickly completed by converting the PNG file to a JPG file.
I was quite surprised about that because I usually use it myself without deleting the background. Can you give me the dataset you used?
and is there a way for us to chat privately on discord or some other way
Or you can message me via https://civitai.com/user/tranthai
desu guen
2024-06-22 14:24:28 +0000 UTC
The effectiveness depends on the quality of the images. When training with images of character, it is best to use software like Photoshop to remove the background.
NFKK
2024-06-22 13:44:30 +0000 UTC
Does the number of photos affect the quality of lora.
or due to base model
Or what method did you use to improve the quality of the output image?
I also tried to create but the output quality is not very good