KREA2 Workflow

I wanted to create a KREA2 workflow tailored to my own requirements and workflow, and I found it useful enough to share with the community.

What it can do

• Enable / disable Character Consistency using KREA2 Identity Edit by conradlocke

• Define image composition using predefined styles

• Standard KREA2 Text-to-Image generation, with or without refinement

• Integrated upscaling

• Eye Detailer and Intimate Area Detailer

• Optional Portrait Master integration

• Visual comparison of the individual workflow stages:

  • Scheduler → Refiner

  • Refiner → Upscaler

  • Upscaler → Detailer

Quality-of-Life Improvements

I've added several features to make the workflow easier and more flexible to use.

Multi LoRA Loader

The workflow includes an integrated Multi LoRA Loader.

You can define trigger words or additional prompts directly for each LoRA. These are saved with the LoRA configuration and automatically merged into your final prompt whenever that LoRA is used.

Prompt Merging

I've also integrated a dedicated text merger and made a small modification to the KSTR Node Pack.

The different prompt components are now merged with a paragraph break between each input, making the final prompt much easier to read and debug.

You can see exactly what is being passed to KREA2 in the Full Merged Text output.

KREA2 Photography Styles

I've created a collection of styles specifically designed to control the visual and photographic characteristics of the generated image.

The styles are therefore not simply collections of keywords. Instead, they are written as KREA2-style instructions that primarily control:

• Camera and lens characteristics

• Photographic aesthetics

• Lighting

• Composition

• Image characteristics

• Era and photographic look

Your actual prompt remains responsible for the subject, person, environment, action and overall scene.

For more information, please check the Notes inside the workflow.

Installation

1. Install the required Custom Nodes

Use Install Missing Custom Nodes in the ComfyUI Manager to install the required node packs.

2. Install the KREA2 Styles

Download:

KREA2 Photography Styles

Copy the file to:

ComfyUI\custom_nodes\comfyui-easy-use\styles

3. Install the modified KSTR prompt node

Download:

prompt_nodes.py

Copy it to:

ComfyUI\custom_nodes\ComfyUI-KSTR-Nodes\

Overwrite the existing prompt_nodes.py.

⚠️ This is a modified version of the original KSTR file. The modification makes the Dynamic Text Concatenate node separate its inputs with a paragraph break.

4. Install KREA2 Identity Edit

Download either or both versions below. I recommend testing both and using only one at a time with KREA2 Identity Edit.

• krea2_identity_edit_v1_1.safetensors

• krea2_identity_edit_v1_2.safetensors

Place the file(s) in your LoRA folder:

ComfyUI\models\loras

5. Install the Detailer Models

Download:

• EyefulV2

• vaginav2.6

Place them in:

ComfyUI\models\ultralytics\bbox

6. Optional: Upscaler

You can use any upscaler you prefer.

For example:

4xClearRealityV1

Place it in:

ComfyUI\models\upscale_models

Recommended KREA2 Setup

As always, you'll need a suitable KREA2 checkpoint.

Some of my favorites:

1. Cielbleu Krea 2

2. Sick Ollie - Krea 2

3. Prime Lust - Krea 2

Text Encoder

I recommend:

Qwen3-VL 4B FP8 Scaled

VAE

As usual:

Qwen Image VAE

A few important notes

Please check the Notes inside the workflow, especially the sections explaining the Styles, Character Consistency and the different workflow stages.

The workflow is designed to be modular, so you can bypass individual sections and use only the parts you need.

I hope you find it useful. Let me know what you think or if you run into any issues. Thanks for trying it!