ComfyUI-TemplateReference

ComfyUI Template Reference

English / 日本語

ComfyUI Template Reference adds two custom nodes for building reusable text, image, and prompt-template libraries directly inside ComfyUI.

Notice

• May 31, 2026 - Update: Added the “Prompt Template LoRA” node, which adds LoRA application functionality to prompt templates.

Features

Template Reference Feature

• Create editable text reference blocks.
• Upload images and preview them as image reference blocks.
• Save and load each block list as a JSON file.
• Select a specific block from the Output dropdown and output the selected text or image to other ComfyUI nodes.
• Drag and reorder blocks.
• Collapse, expand, resize, copy, and delete blocks directly in the node UI.
• JSON data created by the user is saved in ComfyUI’s user directory.

Prompt Template Feature

• Create prompt templates and set a prompt title and prompt text for each block.
• Upload images to each prompt template block and preview them as image references.
• Save and load each block list as a JSON file.
• Select a prompt template from a specific block in the Output dropdown and output the selected prompt to other ComfyUI nodes.
• Drag and reorder blocks.
• Collapse, expand, resize, copy, and delete blocks directly in the node UI.
• JSON data created by the user is saved in ComfyUI’s user directory.

Prompt Template LoRA Application Feature

• In addition to the prompt template feature, LoRA application can be added to LoRA templates.
• JSON data created by the user is saved in ComfyUI’s user directory.
• JSON data saved using the prompt template feature is also loaded and applied.

Nodes

Template Reference

Use this node to collect mixed text and image references.

Outputs:

• select_text: Text from the selected text block.
• select_image: Image from the selected image block.
• template_json: Full serialized node state.

Main controls:

• Open / Save: Load or save the Template Reference JSON library.
• Output: Select the block to send to the output.
• List Off / List On: Turn each block off or on.
• Hide All / Show All: Hide or show all blocks.
• Add Textbox: Add a text reference block.
• Add LoadImage: Add an image reference block.
• Hide / Show: Hide or show the selected block.
• Copy: Copy the contents of the textbox to the clipboard.
• Fit: Adjust the block to fit the size of its text or image.
• Delete: Delete the block.
• Upload: Upload a reference image.

Prompt Template

Use this node to collect reusable prompt templates.

Outputs:

• selected_prompt: Text from the selected prompt block.
• prompt: Same selected prompt text, provided as a convenience output.
• template_json: Full serialized node state.

Main controls:

• Open / Save: Load or save the Prompt Template JSON library.
• none: Output an empty prompt.
• Output: Select the template to send to the text output.
• List Off / List On: Turn each block off or on.
• Hide All / Show All: Hide or show all blocks.
• Add Template: Add a prompt template block.
• Hide / Show: Hide or show the selected block.
• Image: Show or hide the optional reference image area for each prompt block.
• Copy: Copy the contents of the prompt textbox to the clipboard.
• Fit: Adjust the block to fit the size of its text or reference image.
• Delete: Delete the block.
• Upload: Upload a reference image.

Prompt Template LoRA

This node applies LoRA in addition to the “Prompt Template” functionality described above.

Inputs:

• model: Model input
• clip: CLIP input

Outputs:

• selected_prompt: The text of the selected prompt block.
• prompt: The same selected prompt text. This output is provided for convenience.
• lora_model: Model output with LoRA applied
• loraclip: CLIP output with LoRA applied
• template_json: The serialized state of the entire node.

Main operations:

• Open / Save: Loads or saves the JSON library for Prompt Template LoRA or Prompt Template. The filename will have “lora” appended to the end.
• none: Outputs an empty prompt.
• Output: Selects the template to send to the text output.
• List Off / List On: Turns each block off/on.
• Hide All / Show all: Hides/shows each block.
• Add Template: Adds a prompt template block.
• Hide / Show: Hides/shows the target block.
• LoRA: Opens the LoRA screen.
• Image: Shows or hides the optional reference image area for each prompt block.
• Copy: Copies the contents of the prompt text box to the clipboard.
• Fit: Optimizes the block size to fit the text/reference image.
• Delete: Deletes the block.
• Upload: Uploads a reference image.
• LoRA screen: Selects the LoRA model to apply from ComfyUI’s lora folder.
• STRENGTH: Adjusts the strength value of the LoRA influence.

Installation

Clone this repository into your ComfyUI custom_nodes directory:

cd ComfyUI/custom_nodes
git clone https://github.com/Tsubasa109/ComfyUI-TemplateReference.git

Restart ComfyUI after installation.

Storage Location

The extension automatically creates local storage folders under ComfyUI's user directory:

ComfyUI/user/template_reference/prompt_templates/
ComfyUI/user/template_reference/prompt_templates/images/
ComfyUI/user/template_reference/reference_templates/
ComfyUI/user/template_reference/reference_templates/image/
ComfyUI/user/template_reference/prompt_templates_lora/
ComfyUI/user/template_reference/prompt_templates_lora/image/

These folders contain user-created JSON libraries and copied reference images.

Notes

• Reference images are not embedded into JSON files. The JSON stores image metadata and the extension copies uploaded images into its storage folders.
• If a saved image cannot be found, the node keeps the workflow running and returns a small placeholder image while logging a warning.
• Runtime storage folders are ignored by Git and should not be committed to the repository.

File Format

• All JSON template files are saved as UTF-8 without BOM.
• Non-ASCII characters (including Japanese) are stored as-is in the JSON, not as escape sequences.
• When opening saved JSON files in an external editor, make sure the editor is set to UTF-8 encoding.

Requirements

• ComfyUI

License

Apache License 2.0 (see LICENSE)