Version 3.1

What's new in v3.1?

• Fixed the issue with "exploding" the search box when this extension is installed
• Loading of Checkpoints, VAE, Upscalers, and Loras through custom nodes
• Updated workflow to make use of the added node types
• Adjusted the default settings for some parameters in the workflow
• Fixed some reported issues with the workflow and custom nodes
• Prepared the workflow for an upcoming feature

Version 3.0.1

Hopefully the final hotfix for version 3.0 of this workflow

Version 3.0a

Note: The version 3.0 release included an issue that broke HRF, I decided to replace it with this v3.0a hotfix

What's new in v3.0?

• Completely overhauled user interface, now even easier to use than before
• More organized workflow graph - if you want to understand how it is designed "under the hood", it should now be
  easier to figure out what is where and how things are connected
• New settings that help to tweak the generated images without changing the composition
  • Quickly iterate between sharper results and softer results of the same image without changing the composition
    or subject
  • Easily make colors pop where needed, or render a softer image where it fits the mood better
• Three operating modes in ONE workflow
  • text-to-image
  • image-to-image
  • inpainting
• Different prompting modes (5 modes available)
  • Simple - Just cares about a positive and a negative prompt and ignores the additional prompting fields, this
    is great to get started with SDXL, ComfyUI, and this workflow
  • Subject Focus - In this mode the main/secondary prompts are more important than the style prompts
  • Style Focus - In this mode the style prompts are more important than the main/secondary prompts
  • Weighted - In this mode the balance between main/secondary prompts and style prompts can be influenced with
    the style prompt power and negative prompt power option
  • Overlay - In this mode the main*/secondary prompts* and the style prompts are competing with each other
• Greatly improved Hires-Fix - now with more options to influence the results
• A (rather limited for now) alpha test for style templates, this is work in progress and only includes one
  style for now (called test)
• Options to change the intensity of the refiner when used together with the base model,
  separate for main pass and hires-fix pass
• (... many more things probably, since the workflow was almost completely re-made)

Version 2.1

This update to version 2.1 of the custom node extension and workflows contains the latest improvements to the Reborn workflows.

There are some important changes and improvements in this release:

• some bug fixes and improvements to the generated images
• Support for LoRA was added to all workflows, by default the LoRA is set to weight 0.0 so it's effectively disabled unless you raise that value
• Due to popular demand, an implementation of HiRes Fix has been integrated in the text-2-image and image-2-image workflows
• It is now possible to pre-condition the latent image with the refiner, the so-called RBR workflow

Due to the bugfixes that improve the quality of the generated images, the downside to using this version is that images generated with the older version can't be re-created anymore by just loading the old image an re-running the process.

For more information, read the included readme file or the description on the main Github page for this repository.

Version 2.0

This update to version 2.0 of the custom node extension and workflows contains the latest improvements to the Reborn workflow, a new Simple workflow variation of it, a brand new image to image workflow, and a working inpainting workflow.

It also contains some bug fixes, including the first merged PR of this project with a fix for the denoising parameter with the refiner.

For more information, read the included readme file or the description on the main Github page for this repository.

Version 1.0.2

Searge SDXL nodes for ComfyUI
Version 1.0.2

Some new node types to support the improved Searge SDXL Reborn workflow, a completely new SDXL 1.0 workflow created from scratch. A json file for this new workflow is also included in the examples directory.

Version 1.0.1

Searge SDXL nodes for ComfyUI
Version 1.0.1

Some new tool node types were added to make building flexible SDXL workflows a bit easier. The improvements over workflows for version 1.0.0 is that some values can be calculated based on other values, for example the resolution of the upscaled image depends on the resolution of the generated image.

Version 1.0.0

Searge SDXL nodes for ComfyUI
Version 1.0.0

The main node in this release is the SDXL sampler node that utilizes both the SDXL base model and the SDXL refiner model. To make prompting for it easier the second node included in this release is a SDXL CLIP node that takes the CLIP from both the base and refiner and generates the 4 conditionals that the SDXL sampler node expects (base positive, base negative, refiner positive, and refiner negative)