I am in comfyui so I am using highresfix script and upscaler (not always). No loRA's were used to make the image. (okay enchanted eyes the last few ones lol)

It does do yaoi, but don't expect that you can just slap in positions and that it will know. Certain yaoi positions are hard for it to do without details. This is the first checkpoint merge of mine that I've shared.

HighRes-Fix

upscale type: both
pixel upscaler: 4x foolhardy remacri
upscale by: 1.5
hires_steps: 12 - 40 (yes that high depends on sampler)
denoise: .30

V1 & V1.5

Steps: 15 - 20
cfg: 4.5 - 7
Samplers: Euler Karras, Ddim Karras, Uni_pc Karras (those are my fav)

V1.6

More v1 w/ thin lines & less cartoonish like v1.5
Steps: 8 - 20
CFG: 3.5 - 5
Sampler: Euler Karras, Ddim Karras, Uni_pc Karras (those are my fav) & Euler A Normal*

*Euler A Normal gives a more illustrated feel.

I like to use lazypos or lazympos embeddings because I don't want to keep writing best quality, masterpiece, etc. Tho if you do, it changes the image you gen from very flat to a bit more 2d.

Play around with it and have fun.

V1.6 will be the last version.