Marinette Dupain-Cheng from the TV series Miraculous.
The data-set was web-crawled from creative commons NSFW fan arts, handpicked for quality and improved with super resolution Ultra-sharp 4x
The dataset contained 25 images, trained at 10 repetitions for 7 epochs the train took 59 minutes.
Captioning was performed with Interrogator openai and improved manually.
It works best with using values from 0.5 if you want just a resemblance to 0.8 for full character. I have experienced bad results with strength of 0.9 or higher
Description
Marinette Dupain Cheng best on 0.8 - 0.9
FAQ
Comments (9)
Wow perfect model! Can you consider add also chatnoir? That would be awesome!
thanks for the suggestion. I will give it a try
Can you post here some of the creations you made with my model? It would help me a lot to get my work going.
done
1.7 GB for a LORA?
is there a reason why is so heavy?
I guess the reason is empirical, I tried smaller networks and I was not very happy with the results. the network with rank 256 was the only one that produced good results to me for this particular character.
@gmorenomello its the heaviest lora ive seen so far looks good and all but the size really puts me off
Heres hoping you manage to make it lighter 800~MB or so
@Alan_Quake Sorry, this was the first lora I ever done that I was happy with the results. I will have in consideration the size in the future. thanks for the comments.
Hi mate!
Apparently lot of the community is using Pony over the standard SDXL... Can you consider porting this AWESOME model for pony as well?
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.