What all is the set up you use to make you ai pics !!

What engine or generator?

I zoomed I to this and another and the resolution is truly top tier

sealofharps said:
What all is the set up you use to make you ai pics !!

What engine or generator?

I zoomed I to this and another and the resolution is truly top tier

I generate with SD15 checkpoints locally. Workflow for the image above was roughly
Generation:
1. Generate Latent Image at 768x768 with around 50 steps.
2. Add Noise (around 10 to 20% or so)
3. VAE Decode, Upscale, VAE Encode, Cook again at around 920x920 resolution and denoise of 40 to 60%
4. Repeat step 3 with with slightly higher resolution

Final Upscale:
Afterwards I do an upscale using a traditional upscaler. Typically I do Nomos8kSCSCRFormer, Nomos8kSCHAT-L and SSDIR-DAT-GAN and choose whichever version I like most.

In some cases I'll use Lightroom to push contrast a bit, or to remove extra toes or similar (some people are very sensitive to those :D)

Generation Takes about 4-5 min per image, upscaling about 10 min per Image. It's slow, but keep in mind that I'm doing that on a anemic Mobile RTX 2060 with 4 GB of VRAM. Lol.
All of it might sound complex, but it's actually a fairly simple workflow. I'm surprised that it works as well as it does.

Usually the most time consuming task is uploading to E6. I generally make multiple versions of an image and can't seem to decide which one to pick *facepalm*. I have literally hundreds of images waiting to be uploaded.