What has changed?

This is a smaller update from  https://openart.ai/workflows/kodemon/sdxl-koflow-upscaler/xI0NYqjdBn1N63GGqGgI

Re-arranged some items in the workflow and replaced the alpha face mask node with a dependency that was already installed. I was also having some issues with the previous alpha mask node.

I have been playing alot with attempting to get a good styling flow going. But landed on SDXL Prompt Styler + a bunch of notes for constructing prompts. This gives a baseline style with prompting instructions that are more flexible although more manual.

What this workflow does

This workflow aims to provide upscale and face restoration with sharp results. It's mostly an outcome from personal wants and attempting to learn ComfyUI. After trying various out of the box solutions I struggled with generating desired outcomes, with details being washed out and faces being low resolution.

One of the strong suits as of now is the ability to generate pretty decent faces when the actor is further away from the shot.

I have only used it for SDXL so far, but should work with SD1.5 with some tweaking.

How to use this workflow

The upscaler is a simple model upscaler with a range from 0 - 1. By default it's 0.5 which is a good compromise between speed and quality.

You can disable the face rendering with a toggle.

Tips about this workflow

You can lock the seed of the Base Sampler if you want to regenerate the faces without triggering a whole new render pass.

I find the scaling value to best suited to 0.25, .5, 0.75, and 1. I have yet to try 1 as 0.75 takes a while.