Training LoRa necessitate a lot of coffee. If you think that what I do is useful and want to help me keep doing this you can buy me a coffe here: https://ko-fi.com/clumsy_trainer

SD is already capable of generating watercolor images but I was curious to see how far you can push a model to produce good quality waterpaint effect without looking like a filter lazily slapped onto your image in photoshop. Also I thinked that it wouldn't be so hard... and I was soooo wrong. This took like 10/11 failed attempt and it somewhat works but not for everything. I need a broader dataset (atm I used ~100 images) but the images I had from the stocks we use at work were limited in terms of subjects and variety, so

THE MODEL IS A BIT OVERTRAINED

sometimes it won't listen, no matter what, he'll pull off thing that you didn't asked for.

Be patient, this is just for kicks and giggles