very high quality, artist:fensu-san and artist:hioshiru and artist:, rainbow dash, pegasus, pony, solo, looking at you, female, mare, nervous, embarrassed, beautiful, detailed background, ((epic)), high res, bedroom, sugarcube corner
Negative prompt: very low quality, anthro, futa, 3d, sfm, humanized, equestria girls, abstract background, sketch, flexible, gradient background, simple background
Steps: 150, Sampler: Euler a, CFG scale: 18, Seed: 1259011138, Size: 512x512, Model hash: 67ff5385

The difference in resolution is still visible though no matter the size. Compare a native 1080+ resolution picture preview to a 512 one.

I agree, there is something about AI generated images that you can instantly spot most of the time from the thumbnail alone, but not this one personally. I have a hard time believing this is AI generated, but if so, hats off.

@Background Pony #7C9C
It’s not the resolution. As I said, I was able to tell from the thumbnail alone

@Background Pony #7C9C
It definitely does have a style. A lot of it is likely the way the VAE decoder looks along with the kind of errors the base model makes.

@Background Pony #7C9C
Price? Google AUTOMATIC1111. Easy to use / download version. No need to pay anyone anything.

the one art generator I know of offers 256 x 256 for free and 512 x 512 and bigger for a price

@Background Pony #7C9C
It’s the resolution. These are all 512 x 512 cause that’s the resolution it was trained on. If I spent 10x longer on each image making it 2-4x higher resolution it would prob be much harder to tell. My approach atm though is to just carpet bomb and manage a few great pics (that don’t need editing) per 100 or so generated. I or anyone else can always improve them later on though, the seed, prompt and settings are in the images metadata.

It’s weird how one can tell from the thumbnail alone that this is AI generated. There is not much difference between traditional and AI generated, but I guess AI generated art has some kind of style to it.