🎬 Krea 2 Woman Prompt Generator

V3

U HAVE TO READ THIS TO UNDERSTAND IT!⬇️

just put the folder with the .py file in your custom_nodes folder and then search for “Woman Fashion” in comfyUI

The Prompt Generator Now Supports Wildcards

The node can now combine its built-in Python logic with external YAML wildcard files.

This gives you significantly more variety while keeping the prompts structured, readable, and compatible with the node’s existing logic.

A new Wildcards folder is included in the ZIP file. Extract the folder and keep it inside the custom node directory.

How to Use the Wildcards

The node can randomly select one or multiple YAML files and combine their content through the Random YAML option.

This allows you to mix different themes, styles, environments, outfits, poses, effects, and other prompt elements.

For example, combining the 18+ YAML with the Magic YAML can create erotic fantasy scenes involving fire, water, shadows, light, magical clothing, spells, and other supernatural elements.

The sample images shown here were generated primarily with the 18+ YAML. Mixing it with the Magic YAML can produce much more varied results.

Included YAML Categories

The ZIP currently contains:

• A default wildcard collection that is completely SFW

• An 18+ wildcard collection

• A Magic wildcard collection

The default collection is only included inside the ZIP, so make sure to extract it before using the node, if u want a full SFW YAML in ur folder, but it makes things boring. You can also delete any YAML files you don’t want or put them into a .zip archive. However, you always need a YAML file that matches the corresponding prefix. Keep the beginning of the filename unchanged and only add extra words after it (mood.yaml/mood_magic.yalm/mood_whatever_u_want.yaml). In other words, you always need all six YAML files in the end.

Editing the YAML Files

You may freely change, remove, or add entries inside the YAML files.

However, the general structure, category names, and main headings should remain unchanged. The Python code expects this structure and uses it to correctly position the selected elements inside the final prompt.

The YAML format is therefore customizable, but not completely arbitrary.

Future Updates

Future versions will mainly expand the node through additional themed YAML files instead of continuously increasing the size of the Python code.

These files may focus on themes such as fantasy, horror, science fiction, photography styles, locations, cosplay, creatures, magic, or other specialized concepts.

The goal is to eventually let you combine many different themed YAML files through the Random YAML option.

Why Does the Node Use Both Python and Wildcards?

A simple wildcard system can create variety, but it often combines descriptions without understanding where they belong inside a prompt.

You can imagine the Python code as the manager and the wildcards as a group of slightly confused employees.

Every employee brings useful material, but they all communicate differently and do not necessarily know where their work belongs. Without coordination, the final prompt may technically contain everything, but the image model has to guess how the individual parts should be interpreted.

The Python code acts as the coordinating layer. It selects, organizes, connects, and positions the wildcard content so the final prompt remains understandable.

Original Goal of the Node

The original goal was to create a true one-button random prompt generator.

All selected elements should work together through natural language, appear in the correct part of the prompt, and still be capable of producing unusual, chaotic, or unexpected ideas.

The node is not meant to generate completely meaningless randomness. It is designed to create controlled chaos: structured enough for the image model to understand, but unpredictable enough to generate interesting results.

And what is the best way to add even more controlled chaos?

More wildcards, of course.

Here’s one big but, though: the wildcard structure is naturally quite messy, since everything somehow interacts with everything else. So for a beginner or anyone who didn’t create the node, it’s probably difficult to understand what is actually going on.

That’s why I recommend simply enabling both random buttons and starting to generate.

The really, really bad scenario happens when I want to expand the .py node: all previously created YAML files then become obsolete. 🤷🏻‍♂️

Further plans for the future include expanding the available options when pressing the random button and adding a display that shows which file the node has selected. However, that apparently has something to do with Java or something similar, so I still need to look into it. -.-

V2.1
just put the folder with the .py file in your custom_nodes folder and then search for “Woman Fashion” in comfyUI

Hint: The node is still in the alpha phase; I still need to figure out how to add more variety to it. (detail_level: It doesn’t add more detail, it simply adds additional descriptions, such as lighting. Read the description below. The lower the level, the more self freedom the model has to interpret the image.)

New Categories

Double Exposure – completely new, featuring 30 different creative photography techniques, including double exposure with an inner scene, elemental transformation, living statues, light trails, ink dissolution, mirror shards, kaleidoscope effects, constellation silhouettes, glitch art, chalk disintegration, and much more.

New Settings

jewelry_amount (None/Minimal/Balanced/Bold/Extreme) – controls the amount of jewelry, now with more youthful motifs such as hearts, stars, rainbows, and four-leaf clovers instead of chandelier earrings.

expression_override – 14 selectable facial expressions, including Angry, Wink, Tongue Out, Surprised, Laughing, Pouting, Smirking, Duck Face, Confused, Sleepy, Blushing, Cross-Eyed, and more, plus Off and Random.

pose_override – enter custom text for the pose. When left empty, the normal random pose selection is used.

randomize_everything – a toggle that randomizes the category, detail level, camera settings, jewelry, and expression all at once.

detail_level replaces the previous variation_strength and prompt_style fields. It is now a single setting with None, Low, Balanced, and Extreme levels, controlling how many camera techniques, color descriptions, and quality tags are added to the prompt.

camera_view has been completely reworked. It now includes eight “Prefer” options instead of the previous four, with new choices for low-angle views, profile views, three-quarter views, and views from behind. The previous “Only” options have been removed.

More Variety and Less Repetition

All 11 standard categories have received massively expanded settings, with approximately 128 new combinations per category through location × lighting/time combinations.

Eyecatcher received an especially large expansion, with 1,621 possible settings generated through material × arrangement × lighting combinations. Examples include paper cranes, glass shards, and confetti combined with hanging, scattered, or spiral arrangements and warm lighting, cool lighting, or black-and-white photography.

All categories now include twice as many actions and poses, including 43 new dynamic movements such as spinning, running, jumping, and poses with flying hair or fabric, instead of relying mainly on simple standing or leaning poses.

Bug Fixes

The “from above” camera-angle filter was completely broken after being lost during an earlier update. This has now been fixed.

The character was always looking in the same direction because left and right gaze directions were never explicitly defined. There are now 15 clearly specified gaze-angle variations.

The Extreme jewelry setting always mentioned hearts in all four variants. Jewelry motifs are now distributed evenly.

Double exposure and other creative effects were previously too weak or unclear. The templates now declare the visual style first, as recommended for Krea 2, instead of beginning with “photorealistic.”

Other Changes

The automatic makeup description has been removed because it could conflict with character LoRAs. It is now only used in the Cosplay with Makeup category.

The age range has been narrowed to 20–25.

V2

Okay, the node has received a complete overhaul. Version 8.1 had simply become pure chaos, so I rebuilt and restructured almost everything from the ground up.

Several categories have been removed, but an outfit_override option has been added instead. If you leave it empty, the node will randomly choose an outfit. You can also enter any clothing or styling you want, including nudity.

I have also added a jewelry section, allowing you to choose anything from minimal accessories to extremely elaborate jewelry.

As mentioned, the entire node has been heavily reworked and is currently still in a testing phase, so some issues or unexpected results may still occur.

Beast Companion is mainly designed for fantasy creatures and magical animals, while Wild Animal focuses more on real-world animals.

Please don't tell Claude that you're using the outfit override for dirty stuff, he'll get pretty pissed :D

Since I’ve been asked about it quite a few times, I’m finally making the node available here:

I've re-uploaded the file here in English; the old 8.1 version seems to be overwritten by the new one, but that no longer happens with this node here.

🎬 Krea 2 Woman Prompt Generator

The node is currently at version 8.1 and can be freely used and shared.

The LoRA sliders are unfortunately still in German. The node generates natural English prompts for Krea 2 and is mainly designed for images of women between 20 and 25 years old.

Most of the images for my character LoRAs were created with this node, so you can probably get a good idea of what it is capable of.

In the base_prompt field, you can add your LoRA trigger, fixed character details, or anything else you want to keep consistent. This text will always appear before the prompt generated by the node.

I also recommend connecting a Read String node so you can directly see the complete prompt it creates.

Oh, right, it's on a soft nude erotic level

The Avoid Bedroom Cliché option mainly exists because of that lamp. You know the one. That one incredibly persistent, godforsaken bedside lamp that Krea apparently believes must appear in every single bedroom image ever created (But I'm sure that must stem from the logic of the nodes somewhere). This option tries to stop it from spawning next to every bed like some kind of cursed furniture NPC.

The node is not perfect, and I’m definitely not a professional developer. However, I have improved the logic quite a lot, and it can now produce some surprisingly good and natural poses.

The facial expressions can still be a little inappropriate from time to time. Sitting at a peaceful family dinner with your tongue sticking out and looking completely unhinged is probably not always the most fitting choice.

Anyway, have fun with it!

I’ll continue updating and improving the node whenever I feel like working on it or come up with new ideas.