1.1 Update:

-Added more interesting 2 pass PRE upscale algorithm.

-Added LORA\LYCORIS\LOCON etc. loader for 1.5 model.

-You can load and use any 1.x or 2.x models in 1.5 mods.

What you can do:

+Use Original SDXL Workflow to render images.

+Use Modded SDXL where SDXL Refiner works as Img2Img.

+Use SDXL Refiner as Img2Img and feed your pictures.

+Use Modded SDXL where SD1.5 Model works as Refiner.

+Use Modded SDXL where SD1.5 Model works as Base.

+You can load and use any 1.x or 2.x models in 1.5 mods.

+LORA\LYCORIS\LOCON support for 1.5 models in Mods.

+Different Prompt Boxes for Refiner and for Base.

+Disable Base or Refiner models and use just one.

+Change CLIP layer.

+Pre/Post Upscalers.

Main goal of all this workflows is simplicity:

-Minimum nodes and modules (but fully featured).

-Works on 16GB RAM + 12GB VRAM and can render 1920x1920.

-Easy and fast use without extra modules to download.

What you need:

-ComfyUI.

-Pruned SDXL 0.9 Models (Base + Refiner) around 6GB each.

-Close ALL that you don't need in Windows, you need to be around 2GB~.

-Drag and Drop *.json file to ComfyUI window.

About Different Versions:

-Original SDXL - Works as intended, correct CLIP modules with different prompt boxes. Base SDXL model will stop at around 80% of completion (Use TOTAL STEPS and BASE STEPS to control how much noise will go to refiner), left some noise and send it to Refine SDXL Model for completion - this is the way of SDXL.

-Img2Img SDXL Mod - In this workflow SDXL refiner works as a standard img2img model. Base SDXL model will always finish the image as always and saves it, then this image will be sent to refiner, it will add some noise and refine it.

-SD1.5 as Refiner Mod - This will work as Img2Img SDXL Mod, but you can use any 1.5 model as refiner. This is really useful if you want to add some flavor from models you like, also if you are low on RAM or VRAM this is will help you greatly since 1.5 models are small and fast.

-SD1.5 as Base Mod - This will work as Img2Img SDXL Mod, but you can use any 1.5 model as a Base. This is really useful if you want to add some oomph from SDXL refiner for models you like, or just to run some tests.

Tips:

+Don't forget to load VAE for SD1.5 models.

+You can connect and use ESRGAN upscale models (on top) to upscale the end image.

+You can upscale image before refiner kicks in using the R1 two pass upscale module.

+You can connect "Load Image" to R1 and send it to refiner.

+You can disable "For Base" or "For Refiner" loading modules to disable full pipe and not wasting time.

+If you low on RAM or VRAM use "SD1.5 as Refiner Mod" workflow, it is fine.

Here is a cookie for reading all this =D