Step-by-Step Guide Series:
SDXL - TXT to IMG Workflow

This article accompanies this workflow: link

This guide is intended to be as simple as possible, and certain terms will be simplified.

Workflow description :

This workflow allows you to generate images from text. It is optimized to integrate all metadata for simplified online publishing.
It integrates an upscaler/sampler which, although it takes longer than a classic upscaler, allows for much better quality.

Prerequisites :

📂Files :

An SDXL model : like this one
in ComfyUI/models/checkpoints/
An upscale model (optional) : like this one
in ComfyUI/models/upscale_models/
As many LoRAs as you like (optional) : like this one
in ComfyUI/models/loras/
For add detail fonction :
- A BBOX detector model : like this one
  in ComfyUI/models/ultralytics/bbox/
- A SAM model : like this one
  in ComfyUI/models/sams/
- A SEGM detector model : like this one
  in ComfyUI/models/ultralytics/segm/

📦Custom Nodes :

Don't forget to close the workflow and open it again once the nodes have been installed.

Usage :

In this new version of the workflow everything is organized by color:

Green is what you want to create, also called prompt,
Red is what you dont want,
Yellow is all the parameters to adjust the image,
Blue are the model files used by the workflow,
Purple is for LoRA.

We will now see how to use each node:

Write what you want in the “Positive” node :

Write what you dont want in the “Negative” node :

Select image format :

For SDXL it is recommended to start at 1024x1024. Dimensions too far apart can cause problems.

Choose a scheduler and number of steps :

I recommend normal or karras and between 30 and 50. The higher the number, the better the quality, but the longer it takes to get an image.

Choose the guidance level :

I recommend to start at 5. The lower the number, the freer you leave the model. The higher the number, the more the image will resemble what you “strictly” asked for.

Choose a sampler :

I recommend euler or dpmpp_2m. Usually, model designers recommend a sampler in their descriptions.

Define a seed or let comfy generate one:

The seed is the unique identification number of the random starting point of an image. With exactly the same settings (prompt, step, etc.) changing the seed allows you to obtain a new image and not exactly the same one.

Add how many LoRA you want to use, and define it :

LoRAs are sort of "mini models" that are added to a base model. They can be used to create certain characters or styles. If you dont know what is LoRA just dont active any.

Choose your model:

There are many models available on Civitai. Choose the one that best suits what you want to generate.

Activate optional functions : (optional)

Select an upscaler model : (optional)

I personally use RealESRGAN_x4plus.pth.

Select upscale ratio : (optional)

This is how much the resolution of your base image will be multiplied by. Keep in mind that the higher this ratio, the greater the risk of quality loss.

Select an BBOX detector model : (optional)

Here you can choose the detection model. Typically you will use the face or hand detection model.

Now you're ready to create your image.

Just click on the “Queue” button to start:

Once rendering is complete, the image appears in the “image viewer” node.

If you have enabled upscaling, a slider will show the base image and the upscaled version.

Some additional information:

Organization of recordings:

All generated files are stored in comfyui/output/SDXL/YYYY-MM-DD.

Depending on the options chosen you will find:

"YYYY-MM-DD-HH-MM_XXXXXXXXXX_Original" the basic file,
"YYYY-MM-DD-HH-MM_XXXXXXXXXX_Upscale" the upscaled.

SDXL - TXT to IMG

Step-by-Step Guide Series:SDXL - TXT to IMG Workflow