Hey, model looks really impressive :)
I'm using 1.3 on automatic1111, but something is missing the lighting is different and the overall quality feels lower compared to what I've made on sinkin and even trying to replicate one of the samples with the same seed, I don't get the same image.
This is with using the exact same generation data, even using R-ESRGAN General 4xV3 at 25 steps upscaling by 2 and denoising 0.7.
I've checked im not reapplying the vae.
I notice similar behaviour with other models that I use from this site.
Any suggestion for things that I'm not accounting for?