It is really interesting that even faces are very well aligned. What is hardest about creating such LORA? Is it piecing together right dataset?
I'm not really sure how to reply - I think I probably shouldn't have shared these pictures ^^;
I'd kill for that model/lora, too! Unfortunately, my equipment isn't big enough for training models... :,,,(
Hope I'll manage to develop a workflow that works better in the future - I don't want to share the LoRA and then every comment will be about how nobody can get any good results π
I'd kill for that model/lora!!!
LoRA alone won't help very much, you also need to sketch a pose for it and put it under ControlNet/Scribble, and the results are so unstable that you get 6 good pictures out of 600 bad ones D:
wen lora π²
Well I trained a LoRA to get this pose but also had to use ControlNet in Scribble mode as there seems to be no other way to bend this tight XD; (even with a LoRA the bend will be still like a loose U)
Think about AI what you will but these are very promising! They look so good and it's only going to get better... Looking forward to more in the future!
Thank you so much! All these pictures are unedited, just raw results straight from prompt - I could definitely put more time into them but here the only point was to test all kinds of styles and make sure they look equally good :3
What is hardest about creating such LORA? Is it piecing together right dataset?
Unfortunately, my equipment isn't big enough for training models... :,,,(
βοΈ Fold Nugget
____
Like it very much
____
Thank you!
β° Daily Bend
π€ Cold Belt
π§Ί Circus Laundry
π§Ί Circus Laundry
π§Ί Circus Laundry