ComfyUI-OOP Introduction
ComfyUI-OOP is a collection of ComfyUI nodes for building image prompts one visual detail at a time. Instead of writing a long description from scratch, you can choose attributes for a person, animal, location, environment, and art style, then combine them into one prompt.
This approach is useful when you want to revise part of an image concept without rewriting everything else. For example, you can keep a character’s clothing and pose while changing their hairstyle, or keep the location while trying different times of day. Many of the nodes also offer randomize, which can help you explore variations when you do not have a specific detail in mind.
ComfyUI-OOP creates prompts and text-encoder conditioning; it does not provide an image-generation model of its own. Your checkpoint and the rest of your ComfyUI workflow determine how the prompt is used to generate an image.
Demo workflow showing ComfyUI-OOP nodes
How ComfyUI-OOP Works
Think of each OOP node as a card describing one part of a picture. A Hair node can describe hair color and style; a Location node can describe where the scene takes place. The central OOP Node gathers the connected cards into a structured prompt.
The process has three parts:
- Choose details. Select values in the nodes you want to use, such as an eye color, a pose, or a scene mood.
- Connect the sections. Connect the relevant nodes to the OOP Node. Sections you leave disconnected are not included.
- Use the result. The OOP Node produces both the assembled prompt as a
STRINGyou can read and text-encoderCONDITIONINGfor your image-generation workflow.
Each smaller node writes its selected details as labels and values, such as color:Black, style:Short. The OOP Node groups those details under section headings such as Style:, Perspective:, Person:, Animal:, Location:, and Environment:. Person details appear as indented bullet lines beneath Person:.
For example, part of an assembled prompt might look like this:
Style: base_style:PhotoRAW
Perspective: shot:Portrait, angle:Normal, background:Blur
Person:
* Hair: color:Black, style:Short
* Eyes: shape:Round, color:Brown
Location: location:City, object:Statue
Environment: time:Night, mood:FullMoon
You do not have to fill in everything. The first choice in each dropdown is a blank space, meaning “no value.” If you leave a choice blank, its detail is omitted; a section with no selected details can be omitted too. This makes it possible to start with a simple prompt and add detail only where you want more control.
With randomize off, selected values stay the same across runs. With randomize on, supported attributes are selected again on each run. That makes it easier to compare deliberate changes when randomization is off, or discover unexpected combinations when it is on.
ComfyUI-OOP Features
Character and subject details
- OOP Person Node 👤 builds a person description from attributes including gender, age, body shape, ethnicity, hair, eyes, mouth, clothing, and pose. Use it when you want the character’s details grouped together in the final prompt.
- OOP Hair Node 🦰 describes hair color and style, with an optional secondary color. You can, for example, keep a short hairstyle while changing its color.
- OOP Eyes Node 👀 lets you choose eye shape and color, or randomize those attributes when exploring character designs.
- OOP Mouth Node 👄 describes mouth shape, size, and openness. Randomization offers another way to try different descriptions.
- OOP Clothing Node 🥼 describes a top garment and a bottom garment, including their types and optional colors. You can change the outfit without having to rewrite other character details.
- OOP Pose Node 🧍 provides choices for body, hand, and leg poses, with randomization support. It helps you specify what a character is doing rather than relying on a general description.
- OOP Animal Node 🐾 describes an animal’s type, color, and action. For example, it can add a black cat that is lying down to a scene.
Scene and visual presentation
- OOP Location Node 🌍 describes a location and an object within it, with randomization options. A city and a statue, for instance, give the scene a more specific setting than “outdoors.”
- OOP Environment Node ⛅ sets the time of day and scene mood, with randomization support. Changing these choices lets you try a different atmosphere while keeping the same subject and location.
- OOP Style Node 🎨 combines visual styles, with choices such as Abstract, Vintage, Cyberpunk, and Watercolor. Use it to describe the artistic direction you want the image generator to follow.
- OOP View Node 📐 specifies camera angle, shot type, and background blur. A portrait shot with a blurred background, for example, describes a different composition from a wider view.
Prompt assembly and encoding
OOP Node 💫 brings the connected sections together. It outputs the readable prompt as a STRING and encodes that prompt as CONDITIONING using the text encoder supplied by your checkpoint workflow. You can connect the text-encoder output from a compatible checkpoint, including workflows using SDXL, SD1.5, Flux, Flux2, Flux3, Krea2, Ernie, or another model with a suitable connected encoder.
The node’s clip_skip setting defaults to 0, which keeps the layer selected when the checkpoint was loaded. For example, an SDXL checkpoint may already be loaded with a -2 setting. You generally do not need to change clip_skip; negative values are for re-encoding with a different layer on CLIP-style encoders.
ComfyUI-OOP Example Workflow
The repository includes a saved example workflow. It is a starting point for seeing how the OOP nodes fit into a larger image-generation graph. The OOP nodes themselves do not need extra node packages, but the saved example uses two nodes from other packages:
| Node used in the example | Where it comes from | ComfyUI alternative |
|---|---|---|
SDXL Quick Empty Latent (WLSH) | wallish77/wlsh_nodes | Empty Latent Image |
| `ShowText | pysssss` | comfyui-custom-scripts in the Comfy Registry |
The example also uses standard ComfyUI nodes: CheckpointLoaderSimple, CLIP Text Encode, KSampler, VAEDecode, PreviewImage, and Conditioning Zero Out.
Troubleshooting ComfyUI-OOP
My selected detail is missing from the prompt
- Check that the node containing the detail is connected to the appropriate section of the OOP workflow.
- Check the detail’s dropdown. A blank-space choice means “no value,” so it will not appear in the prompt.
- If an entire section is absent, check that it contains a selected value and is connected to the OOP Node.
My character or scene changes between runs
Check whether randomize is enabled on any of the nodes describing that subject or scene. Turn it off for attributes you want to keep steady.
The saved example workflow reports missing nodes
The example graph uses nodes outside this package. Check the missing node’s name against the example workflow table. You can obtain the listed package or substitute the indicated ComfyUI alternative. Missing example-workflow nodes do not mean the OOP nodes themselves require those extras.
I cannot see how my choices became a prompt
Use the OOP Node’s raw STRING output to inspect the assembled text. Its section labels and bullet lines make it easier to spot a missing connection or an unwanted blank choice.
Frequently asked questions
Do I need to use every OOP node?
No. Every section is optional. Connect only the parts you want in your prompt.
Does ComfyUI-OOP include a checkpoint or image-generation model?
No. It organizes and encodes prompt text using the text encoder connected from your checkpoint workflow.
Should I change clip_skip?
Usually not. The default value, 0, retains the layer chosen when the checkpoint was loaded. Negative values are intended for changing the encoding layer on CLIP-style encoders.
Can I keep some details fixed while exploring others?
Yes. Keep randomize off for the details you want to preserve, and enable it on supported nodes whose attributes you want to redraw.
Learn More about ComfyUI-OOP
- Explore the ComfyUI-OOP repository for the project files and documentation.
- Open the demo workflow image to see the nodes arranged visually.
- Examine the saved example workflow for a fuller example of how the prompt-building nodes connect to an image-generation workflow.
