Basic model:Anything diffusion v4.5(!!!Vae Not Nedeed!!!)
Images were taken here:https://twitter.com/scott_malin
don't forget to send in your results
P.S. I opened a discord server where you can offer your ideas for models, link:https://discord.gg/HDfvejBPMJ
Description
it's more like his style now
FAQ
Comments (5)
Guess someone was faster than me :)
I can see a bit of Scott Malin, but it still lacks the "cartoony" style. It feels more like you can kind of see Lucia's face but that's it;
This might need some more work, also, consider making a LoRa as well !
With Lora I have a little problem in terms of understanding and quality, I do not even know how many pictures you need to teach her
@ANONIM4IK3327 You don't need many. Personally I trained a model on Bacun's art and used around 80 images in the training (worked pretty well, though I'm still improving the dataset).
I didn't have any idea of what I was doing, so I just followed AItrepreneur's and Kasukanra's guides.
I'd say you have more than enough here, starting at 45:15 : https://www.youtube.com/watch?v=7m522D01mh0
what training method do you use? dreambooth or finetune.?
dreambooth