This is a versatile exclusive workflow for Krea2, which can fully replace the default Krea2 workflow in ComfyUI. It delivers superior quality, offers more flexible operation and stronger compatibility. It works exceptionally well for stacking stylized LoRAs while barely altering the image composition. I will walk you through how to use it below.

First, here is what it looks like

The workflow uses a self-developed node of mine: NO8D-Controls。Get it here.

Alternatively, search for NO8D-Controls in ComfyUI Manager

This is a two-stage workflow. The first stage can take Klein as the base model, allowing you to access Klein's extensive LoRA library, such as our slider-LORA series.

Images generated via Klein are fed into the second-stage Krea2 pipeline to upscale, enhance the visuals and add intricate details.

In addition, I have set up node groups for prompt expansion and image inversion to make it easier for you to fill in prompts, and they are located on the far left side of the workflow.

The node group on the far right of the workflow contains supplementary optional settings, which are not mandatory. It includes noise addition, RTX upscaling, and save settings.

When you open the workflow, you may notice that the Klein node group and the Krea2 node group share highly similar compositions. Indeed, they are interchangeable.

The advantage of this approach is that you can use this workflow to optimize and fix the outputs generated by Klein, especially for limb correction. Note that the Denoise value should preferably be between 0.3 and 0.6, which delivers richer details without causing excessive distortion.

Through two-stage generation, the final images also gain more abundant details. However, judging from the generation results, Klein can deliver more diverse visual variations, so I set Klein as the default option.

Replace the model within the Klein node group with Krea2, and the two generation stages will share the same base model. As a result, perfectly identical images can be generated with the same seed value. Leveraging this feature, you can significantly increase the denoising value to obtain far richer image details. If you load a style LoRA model at this point, you will find that this workflow transforms into a style transfer workflow.

This is the rendering result when the denoising strength is set to 0.9, with the LoRA model titled atmospheric photography applied. It is obvious that the light and shadow effects of the image have been greatly enhanced, while the image content remains almost unchanged. This is the most powerful advantage of this workflow: it is fully compatible with all your Klein-version LoRAs. It can serve as an upscaling workflow by adding details via Krea2, and can also flexibly integrate LoRA models to function as a style editing workflow without altering the original image itself.

OK! That's all. I hope you'll like it.

🫡 May the AI-power be with you, see you soon !🫡

WqcTN_yzuTbnoMKPvI5SR8H1kHIRRfb0RValVDoFafQ

Click here

Disclaimer: My model or workflow is solely for learning and communication purposes and shall not be used for any commercial activities. All responsibilities for the generated content, including its legality, ethical implications, or potential misuse, are borne entirely by the user. By using this model or workflow, you agree to this principle. Please use it responsibly within the scope of local laws.