LoRA is generated entirely from manhwa images.
-----------------------------------
Note:
Recommended neg.: 'shaded face, loli'
Check samples for prompts :)
-----------------------------------
Leave a like, comment, review but first of all, ENJOY ;)
Description
FAQ
Comments (10)
What upscale model should I use for Hires fix?
I am using Latent since nothing else is specified, but the resulting image is very fuzzy. (Aside from the face surprisingly enough)
I don't use upscaling for Pony at all, only adetailer for face.
No idea what can be causing a fuzzy image on your side, try a different sampler.
@LittleJelly weird. because when i copied your generation settings from the first image you posted, it enabled hires fix by itself.
I tried multiple samplers, and only R-ESRGAN 4x does not have the fuzziness, but the face quality is lower with that one.
I'll try adetailer next
@ralkey When you use PNG info in a111 it always automatically throws hires fix. Idk why's that.
I get quite harsh black lines in my generation, but when I put the generation resolution to my native resolution, the black lines are less harsh.
However, having such a high resolution makes it painfully slow to generate.
What did you do to avoid the harsh black lines?
What resolution do you use?
@LittleJelly I set the resolution to 984x1440, which is much closer to me screen res.
It does mostly get rid of the harsher black lines, but its slow to generate.
@ralkey that's not an Pony supported resolution if im not mistaken, lower it down to: 832x1216, that's the resolution i always generate at.
Can you make SMLN Lora's for IL please?
I am wanting to completely convert to IL due to its better prompt adherence and overall better quality,
However, I use your SMLN Lora's the most and they're not compatible.
You are not the first one to ask me, so I'll move it up on my ToDo list.
However I won't be able to make it within the next week due to full IRL schedule.