Lora of Guardian tales chibi style
[English] (ChatGPT translate)
The lora work well between 0.6 and 1, but i recomend to set it to 0.7.
Differeces between v1.0 and v640x640
v640x640 was trained with the same dataset but with a resolution of 640x640. I did this mainly to test if it could solve some issues with the original images, as lowering the resolution fixes some defects such as improperly painted parts and less perfect edges.
The result wasn't very good XD. But for some reason, the images turned out somewhat blurry and adopted a style similar to being drawn with crayons, that was interesting to me, so I decided to upload the model anyway
Not useful information
I am starting to practice how to create Loras in XL. This was the result of the first tests I did, so the results may not be as good. Now I am going to recreate my previous models but in PonyXL.
i don't know how to make images using PonyXL, so i put the first thing that the ia generate, i would be grateful if you could post images of the results
What does the original art look like?
If you're not familiar with the drawing style of Guardian Tales chibis, here's an example of how they look.
Description
Lora trained with the model PonyXL
images = 54
Resolution = 1024x1024
Problems:
Sometimes the eyelashes don't look very good
In models that are not very good at creating chibi characters, the faces may not look as good
FAQ
Comments (5)
Can i have your dataset and train settings? Your lora performance is GREAT!
I uploaded the dataset in the files section. I don't have the .json file because I trained the model using Google Colab, but you can still check the metadata of the file to see the training settings
Love this!!
Why can't it be used in tensorart?
The file got corrupted, so I re-uploaded it to TensorArt, and it seems to be working fine now.
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.