The Wildcard Manager is a powerful dynamic prompt and wildcard processor. It allows you to create complex, randomized text prompts using a flexible syntax that supports nesting, weights, multi-selection, and more. It is designed to be compatible with the popular syntax used in the Impact Pack's Wildcard processor, making it easy to adopt existing prompts and wildcards. It supports both .txt ( default one, not needing the extension) and YAML files ( both .yaml and .yml)

Reading the files from the default ComfyUI folder ( ComfyUi/Wildcards )

✨ Key Features & Syntax

Dynamic Prompts: Randomly select one item from a list.
- Example: {blue|red|green} will randomly become blue, red, or green.
Wildcards: Randomly select a line from a .txt file in your ComfyUI/wildcards directory.
- Example: __person__ will pull a random line from person.txt.
Nesting: Combine syntaxes for complex results.
- Example: {a|{b|__c__}}
Randomize: Randomize the wildcard independently from the seed.
- Example: __*person__ w ill always pulls a random line from person.txt, regardless of the seed.
- Example: {*red|blue|green} always pulls a random item from the list, regardless of the seed.
Sequential: Use + to select items in order across different queue runs.
- Example Wildcard: __+person__
- Example Dynamic: {+red|blue|green}
Weighted Choices: Give certain options a higher chance of being selected.
- Example: {5::red|2::green|blue} (red is most likely, blue is least).
Multi-Select: Select multiple items from a list, with a custom separator.
- Example: {1-2$$ and $$cat|dog|bird} could become cat, dog, bird, cat and dog, cat and bird, or dog and bird.
Quantifiers: Repeat a wildcard multiple times to create a list for multi-selection.
- Example: {2$$, $$3#__colors__} expands to select 2 items from __colors__|__colors__|__colors__.
Comments: Lines starting with # are ignored, both in the node's text field and within wildcard files.

🔧 Wildcard Manager Inputs

wildcards_list: A dropdown of your available wildcard files. Selecting one inserts its tag (e.g., __person__) into the text.
processing_mode:
- line by line: Treats each line as a separate prompt for batch processing.
- entire text as one: Processes the entire text block as a single prompt, preserving paragraphs.

🗂️ File Management

The node includes buttons for managing your wildcard files directly from the ComfyUI interface, eliminating the need to manually edit text files.

Insert Selected: Insertes the selected wildcard to the text.
Edit/Create Wildcard: Opens the content of the wildcard currently selected in the dropdown in an editor, allowing you to make changes and save/create them.
- You need to have the [Create New] selected in the wildcards_list dropdown
Delete Selected: Asks for confirmation and then permanently deletes the wildcard file selected in the dropdown.

📸 Screenshot / Demo

Wildcard - Entire Text As One
Wildcard - Line By Line
Wildcard - Comptability with PromptList for a batch processing
Wildcard - View Content

LoRAs Nodes

🎲 Randomize LoRAs Node

The Randomize LoRAs node lets you experiment by randomly selecting LoRAs from a list and assigning randomized strengths. You can enable exclusive selection (just one LoRA) or allow multiple LoRAs to be mixed. There are two nodes for this, the Random LoRA Selector and the Random LoRA Selector to Model Selector, the difference between them is that one uses the lora_stack and the other one applies the loras directly to the model.

🎲 Random LoRA Folder Selector

The Random LoRA Folder Selector node lets you experiment by randomly selecting LoRAs from a doler and assigning randomized strengths. You can enable exclusive selection (just one LoRA) or allow multiple LoRAs to be mixed and how many to be selected by folder. Additionally, if the LoRA includes trigger words (via lora-info), these are extracted and returned for use in prompts. There are two nodes for this, the Random LoRA Folder Selector and the Random LoRA Folder to Model Selector, the difference between them is that one uses the lora_stack and the other one applies the loras directly to the model.

✅ Compatible with comfyui_image_metadata_extension by using the 💊 CR Apply LoRA Stack from Suzie1 / Comfyroll_CustomNodes

✅ Compatible with comfyui_image_metadata_extension by using the LoRA Metadata Hub

🎲🚫 Excluded Loras

The Excluded Loras node lets you select 5 lroas to be excluded in the Random LoRA Folder Selecto, this way you don't need to worry about the select getting an unwanted lora from the folder

🎲⛏️ Extract And Apply LoRAs

The Extract And Apply LoRAs node lets you extract the LoRAs that were saved using the comfyui_image_metadata_extension node and apply them to a workflow. This was created so I could try to generate better enhancement workflow but with the same loras and weight as the initial one. I would recommend use the SDBatchLoader to load multiple images at the same time and connecting the output to the path field in my node

🎲📦 LoRA Cache Preloader

This is a node to have the information for the loras preloaded into the db.json You can select the folder that you want to run it. This way you don't need to wait for the information to be gathered when running the other two nodes

🔧 Node Inputs

I'll describe only the inputs that aren't straight forwards here:

refresh_loras: To force a refresh of LoRAs everytime
exclusive_mode:
- On – choose 1 LoRA only
- Off – choose a random number from the list
stride: Number of runs before re-randomizing (not currently enforced)
lora_count: the number of LoRAs to be extracted from the list ( if 0 it will be a random value / exclusive_mode will overcome this is set to On )
lora_stack: (Optional) Input for chaining or merging with existing LoRA stack
extra_trigger_words: (Optional) Additional text to append from previous node
force_refresh_cache: Set to True to clear and regenerate cached LoRA metadata

📤 Node Outputs

LORA_STACK: A list of (name, min_strength, max_strength) tuples
Trigger Words: Comma-separated string of trigger words for selected LoRAs
Help Text: Describes the meaning of each setting for reference

📤 Cache Behavior

LoRA metadata (like trigger words) is cached to speed up repeat runs.
Use force_refresh_cache: True to clear and reload cache.
This helps avoid performance issues with large LoRA libraries.

📸 Screenshot / Demo

preview

🔄 Other Nodes:

PromptList w/ Template

Based on the PromptList from Impact, but this one will be able to save, edit and delete templates

SaveWorkflowAndShutdown

A simple node to shutdown the PC at the end of the workflow if there are no more jobs in the queue. There is a delay input, if you want to interrupt, cancel comfyui's process

Split Batch With Prefix

A node to to transform a batch of images into a list. It will also ask for a prefix and an index, so it will generate a prompt list. The purpose is to save each image from a batch with a corresponding index. Example, you have a 5 images batch, you run the node and will output the images as a list and a prompt string of the index+Prefix ( example 0-SDXL_, 1-SDXL_, 2-SDXL_, 3-SDXL_, 4-SDXL_)

Model Assembler

A Model to combine the model checkpoint, unet loader, clip loader and vae loader. Created so I don't have to change connections when testing full models and fp8 versions

LoRA Metadata Hub

Node used to add the metadata from the loras to the image metadata together with the comfyui_image_metadata_extension node. You may need to add/edit the modules/defs/ext/SantodanNodes.py file to add the following code - SantodanNodes.py - since the node OP hasn't updated the node for a couple of months, or just my whole node

📥 Installation

Option 1: ComfyUI-Manager (recommended)

Open ComfyUI
Open ComfyUI-Manager
Search for Santodan or Randomize LoRAs
Click Install

Option 2: Manual Install

cd ComfyUI/custom_nodes
git clone https://github.com/Santodan/santodan-custom-nodes-comfyui/

Then restart ComfyUI.

🔁 Updating

cd ComfyUI/custom_nodes/santodan-custom-nodes-comfyui
git pull

Restart ComfyUI to apply the update.

❌ Uninstallation

cd ComfyUI/custom_nodes
rm -rf santodan-custom-nodes-comfyui

Restart ComfyUI.

🧠 Credits

Suzie1 / Comfyroll_CustomNodes
Provided the original randomization logic inspiration.
jitcoder / lora-info
Used for extracting trigger words from LoRA files.
ChatGPT
Used to create the whole code as I only have some basic programming knowledge
Claude.ai
Used to create the whole code as I only have some basic programming knowledge
Gemini (Google AI)
Used to create the whole code as I only have some basic programming knowledge