Quoted By:
Full Disclosure: I am an idiot most of the time
After updating the Thread Gallery with the most recent threads I realized "Wait a minute! I have all these nice pics on hand that anons worked really hard to make and most are of Masterpiece quality..". So I got to work.
I used the Aesthetic tagger to sort the images from just 3 of some old threads which got rid of screenshots and Skynet logos as "worst quality" (sorry) and kept only good and above. Then, I trained a test lora on about 400 images.
The result is this.
I wasn't surprised it learned better about how Fauna looks as I tagged the images with Character tags on and there were a bunch of my Faunas in the dataset. Other than that, lines were thicker and there was more contrast.
Part 1/2
