- - Requested Character
Buy me coffee here: https://ko-fi.com/baipaiyok or https://buymeacoffee.com/baipaiyok
——————————————
V.3-PonyXL & V.3-Illustrious
Fixed:
quality of dataset
strange shadow drop
prompt: bishopRO, long sleeves, blue dress, puffy sleeves, cleavage cutout, shrug (clothing), yellow sash, side slit, fishnet, white thighhighs, black high heels
V.2-PonyXL & V.2-Illustrious
Trust me, Illustrious always better than Pony Lol.
Fixed:
Reduce File size
More Flexible for apply style
More accurate prompt
prompt: bishopRO, white thighhighs, long sleeves, blue dress, puffy sleeves, cleavage cutout, side slit, shrug (clothing), yellow sash, black high heels
V1.0-Illustrious
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://civarchive.com/models/795765/illustrious-xl?modelVersionId=889818
Recommended checkpoint ----> https://civarchive.com/models/827184?modelVersionId=1025184
prompt: bishopRO, white thighhighs, fishnet legwear, long sleeves, blue dress, puffy sleeves, cleavage cutout, side slit, jewelry, cross, shrug (clothing), yellow sash
additional: black high heels
V1.0-PonyXL
Please Read
This V1.0 have 2 files.
V1.0-more accurate = Have more accurate details
V1.0-more flexible = I have low the learning rate to make it more flexible
This is my first LoRa by PonyXL. Feel free to try.
I will update other LoRa to PonyXL soon.
Prompt: bishopRO, white thighhighs, fishnet legwear, long sleeves, blue dress, puffy sleeves, cleavage cutout, side slit, jewelry, cross, shrug (clothing), yellow sash
Description
Just init my first PonyXL
FAQ
Comments (6)
Wow, this is really accurate. Incredible work as usual! The only thing is it seems to change the style/bleed into the original model's style when set at a high LORA strength. Hopefully can be improved in the future, but this is already super good!
Thank for review. I'll fix it in a future. I guessed my accuracy rate too high for this case.
@baipaiyok I think it's already super good, but yeah perhaps in the future when PonyXL v7 is out haha (if ever)
@donaldoctavia22611 I have a same issue while working on Acolyte LoRa right now. The output look bad because it looks like images in a dataset. Lol I must fixed it because SD 1.5 dataset a bit low quality.
Any chance for a version on Illustrious? Kind of want to try to see if there is a difference in accuracy :)
@donaldoctavia22611 ok I'll make it one