About the workflow

Latest update:

Introduced a new upscaler that splits the images into different parts, upscales each one individually and ads details. After that it merges the images and creates a bigger one. Optionally, it can be scaled again and another sampling can be done to further enhance.

I created a workflow for ComfyUI that I believe has a few advantages compared to others. The main features are:

1. Works with SDXL, SDXL Turbo as well as earlier version like SD1.5

2. Has flow for splitting the image into multiple parts, upscaling and adding details and merging them to create a bigger, more detailed image

3. Has 5 parameters which will allow you to easily change the prompt and experiment

4. Toggle if the seed should be included in the file name or not

5. Upscale to 2x and 4x in multi-steps, both with and without sampler (all images are saved)

6. Details and bad-hands LORAs loaded

I use it with dreamshaperXL mostly and works like a charm.

How to use

The workflow has multiple groups for prompting, the actual image generation, two upscalers and a configuration for the file name.

Prompting

Use the positive and negative prompts just like in any other workflow. The advantage is that there are up to 5 parameters that you can use to easily change the prompt. In the provided example you can change the color, nationality or age with ease.

Upscale

There is an upscale that will upscale the image 2 times, both using just hiResFix and with another sampler. Both images are saved.

A secondary upscale will upscale again 2 times the resulted image if enabled.

By default, both upscales are disabled.

File name info

You can easily change the file name and if the seed should be included in the name as well.