After I studied how to make it better. Finally I got it. Prodigy not good for character like Ragnarok Online. The output is overfitting.
Special Thank to ChameleonAI and Leonmitchelli
Fixed:
Reduce File size
More Flexible for apply style
More accurate prompt
prompt: swdro, breastplate, brown capelet, dress, belt, armored gloves, long skirt, brown skirt, gauntlet, yellow gloves, frills
*****-- Remove "frills" If it show to much
prompt: swdro, breastplate, brown capelet, dress, belt, armored gloves, long skirt, brown skirt, gauntlet, yellow gloves, frills
*****-- Remove "frills" If it show to much
I use a same dataset as PonyXL. I didn't test it much. Feel free to try.
Please request a LoRa you want to create to Illustrious.
My base for training ----> https://civitai.com/models/795765/illustrious-xl?modelVersionId=889818
Recommended checkpoint ----> https://civitai.com/models/827184?modelVersionId=1025184
prompt: swdro, armor, capelet, armored dress, belt, armored gloves, long brown skirt
additional: floral print // In case the flower print on shoulder not show.
This LoRa have some problem when do pose sitting on floor. Its may not show long enough skirt. Feel free to try.
prompt: swdro, armor, capelet, armored dress, belt, armored gloves, long brown skirt
additional: floral print // In case the flower print on shoulder not show.
weight: 0.7-1.0
This LoRa created from very small dataset. It may not show a correct outfit with some pose.
prompt: swdro, armor, capelet, armored dress, belt, armored gloves, brown skirt
This is my first Lora. Feel free to use.
------------------- V.2
V.1 more flexible than V.2 but V.2 have more consist than V.1
promtp : swdro,boots, armor, yellow skirt, brown dress, long dress, flora print, rose print, armored dress, boots
additional : holding sword, holding weapon, belt
------------------- V.1
Recommended weight: 0.6 to 0.8
trigger prompt : ,swdro, 1girl, armor, belt, boots, dress,
for weapon : holding sword, sword, weapon,
***The result that included weapon prompt may cause abnormal weapon in an image.
V1.0
made from dataset contains 15 images
trained with 240 steps
weapon result may look abnormal. please use other Lora weapon to adjust for example : https://civitai.com/models/145611/conceptswordovershoulder
Go ahead and upload yours!
Your query returned no results β please try removing some filters or trying a different term.