Eugene's ComfyUI Custom Utility Nodes

A collection of utility nodes designed to optimize and manage workflows in ComfyUI.

Dictionary-Based Prompt Management

Utility nodes for systematically managing and reusing prompt texts using Python Dictionaries.

Dictionary Nodes

• DictUpdate1

  • Adds/updates a single key-value pair in a dictionary.
  • Creates a new dictionary if no input dictionary is provided.
  • Ignores empty keys or values.

• DictUpdate5

  • Adds/updates up to 5 key-value pairs simultaneously.
  • Creates a new dictionary if no input dictionary is provided.
  • Automatically ignores empty keys or values.
  • All fields are optional.

• DictUpdate10

  • Adds/updates up to 10 key-value pairs simultaneously.
  • Creates a new dictionary if no input dictionary is provided.
  • Automatically ignores empty keys or values.
  • All fields are optional.

• DictTemplate

  • Applies dictionary values to a template text.
  • Supports multi-line text.
  • Syntax: placeholders are defined as {key}.
  • Example:

• DictMultilineSelect
  • Selects specific lines from multi-line text and adds them to a dictionary.
  • Input:
    • Existing dictionary.
    • Line number to select.
    • Multi-line text.
    • New key string.
  • Output: Updated dictionary and selected line number.

Dictionary Bus Nodes

Utility nodes for managing dictionaries alongside model components (MODEL, CLIP, VAE, etc.).

• DictBus

  • Bundles a dictionary with model components into a single unit.
  • Required inputs: dictionary, model, clip, vae.
  • Optional inputs: image, latent.
  • Allows multiple components to be passed as a single connection in workflows.

• DictBusUnpack

  • Separates a DictBus bundle into individual components.
  • Outputs all components (dictionary, model, clip, vae, image, latent) individually.
  • The original bus is also outputted for continued use.

• DictBusEdit

  • Selectively modifies specific components of a DictBus bundle.
  • Unmodified components retain their original values.
  • Allows for updating specific components mid-workflow.

LoRA Preset Management

Nodes for managing and applying LoRA models and prompts as presets.

Preset Management Nodes

• LoraPresetSaver

  • Saves up to 5 LoRA settings as a preset file.
  • Configurable options for each LoRA:
    • LoRA model selection.
    • Strength adjustment.
    • CLIP strength adjustment.
    • Positive/negative prompts.
    • Alias (nickname) assignment.
  • Presets are saved in JSON format.

• LoraPresetSelector

  • Selects and manages saved LoRA presets as a list.
  • Supports a subfolder-based preset structure.
  • Provides real-time preset refresh functionality.
  • Offers dynamic preset management via a web interface.
  • Bypass option for selective application.
  • Maintains a list of selected presets.

LoRA Loader Nodes

• Lora Preset Loader & Encoder
  • Applies up to 5 LoRA presets simultaneously.
  • Allows individual adjustment of each LoRA's strength.
  • Supports prompt prefixes/suffixes.
  • Compatible with Dictionary-based prompt templates.
  • Allows CLIP layer adjustments.
  • Works with Dictionary Bus and Basic Pipe.

• **Lora Preset List Loader & Encoder **
  • Directly uses the output list from PresetSelector.
  • Sequentially applies multiple LoRAs.
  • Uses settings saved in the preset.
  • Compatible with Dictionary Bus and Basic Pipe.
  • Supports prompt prefixes/suffixes.

NOTE: Override

When multiple inputs are connected to a node, the following override rules are applied:

1. Priority 1: Direct input to the node.
2. Priority 2: Input from basic_pipe.
3. Priority 3: Input from dict_bus.

Example

If the model is connected to the node via dict_bus, basic_pipe, and directly to the node, the model directly connected to the node will be used.

Installation

1. Clone this repository into the custom_nodes directory of ComfyUI.
2. Restart ComfyUI.

Additional Information

Web Interface

Provides web API endpoints for managing LoRA presets:

• /lora_presets: Retrieves information about all available presets.
• /lora_presets/refresh: Forces a refresh of the preset list.

Preset Directory Structure

Presets are managed within the LoRA folder in the following structure:

loras/
├── preset1_preset.json
├── category1/
│   ├── preset2_preset.json
│   └── preset3_preset.json
└── category2/
    └── preset4_preset.json

Preset File Structure

{
    "lora_name": "model_name.safetensors",
    "lora_path": "/path/to/lora",
    "strength": 1.0,
    "clip_strength": 1.0,
    "prompt_positive": "positive prompt text",
    "prompt_negative": "negative prompt text",
    "nickname": "Display Name"
}

Usage Examples

• Workflows: Refer to the examples in the ./workflows directory.
• Presets: Presets are generated automatically, but you can also refer to ./Sci-fi_Enviroments_preset.json for inspiration.