It's one of my first attempts to train in SDXL. It's some kind of experiment, but I want to publish this one as results are quite cool in my opinion. There is of course enough space to improvements, but getting in consideration that images generated are only with SDXL 1.0 without refiner, I am happy by now with generations.
LoRA weights should be between 1.1 and 1.4 in my experience.
For portraits you can use 896 x 1152, 832 x 1216, 768 x 1344, 640 x 1536 or even 768x1024 resolutions.
This LoRA has been trained after carefully selecting 31 images. All photos have been enhanced to make the most of them. I hope you like it!
I don't ask for anything but any respectful constructive review (after having tested the model) would be highly appreciated or at least press Like if you're happy with results after trying it out! Thanks in advance and enjoy it!
Notice: This free work is intended for art and illustration purposes of emergent technology and not intended nor' trained for making illicit images, as the author I've taken care to not use any private data in the training process or designed it for such purposes. By downloading this model you are agreeing with the aforementioned intent and that you will not use my work to create / share images of a potentially immoral and / or illegal nature.
Description
FAQ
Comments (7)
Can I ask how and where you train your sdxl lora? thank you
Of course! I use kohya_ss for making all my models! It is similar to training SD 1.5 but with other parameters! I'm still experimenting though!
@astragartist care to share the current parameters you used to train this model?
@sniper1999 It's very easy! Just download model on automatic1111, put into lora folder. goto lora tabs amd press "i" on it and you get full information. ;)
the left eye is totally wonky. May be wrong but I think that might be to do with overtraining.
It could be! Thanks for advise! I'm still in trial and error but in right direction!
A good first step, needs a little more training or data, I think