Try to generate any higher and you’ll get very weird and ugly results.
Stable Diffusion in particular has this issue with limbs. People have 2 arms at 512x512? Surely they must have 4 arms at 1024x1024! That’s just math.
Try to generate any higher and you’ll get very weird and ugly results.
Stable Diffusion in particular has this issue with limbs. People have 2 arms at 512x512? Surely they must have 4 arms at 1024x1024! That’s just math.
rufus@discuss.tchncs.de 10 months ago
Yeah, It’s the wrong way to do it and will lead to unusable results. The correct way to do it is to generate an image with the resolution the AI was trained on an then upscale it.