ComfyUI-Meld is a suite of high-performance infrastructure nodes designed to streamline complex ComfyUI workflows. It aggregates multi-step logic--such as advanced prompt construction, metadata extraction, and adaptive image processing--into single, atomic operations.

Designed for stability, security, and zero-dependency hell.

Key Features

Workflow De-cluttering: Replaces dozens of primitive nodes with a single "Meld" node, significantly reducing canvas noise.
Zero Dependency Hell: Works out-of-the-box using standard libraries. No external pip install required.
Privacy-Focused Metadata Handling: Extracts only generation data, naturally ignoring private EXIF fields.
Scoped File Loading: Reads files exclusively from the specified directory and its subdirectories.
Advanced Prompting: A powerful engine that handles file-based wildcards, dynamic weights, and automatic negative prompt extraction.
Meld Image Manager: Integrated image management system with search, tagging, and lineage tracking.

Node Usage: Meld Save Image

This node saves generated images to the output directory and automatically registers them in the Meld Image Manager database.

Automatic Metadata Extraction: Saves prompts, workflow, and model information.
Lineage Tracking: Automatically links child images to their parents based on metadata or pHash.
Tagging: Support for adding custom tags at generation time.

This node is the heart of workflow organization. It allows for dynamic prompt generation with a sophisticated randomization engine.

Random Prompt Syntax

You can use the following syntax directly in the text input or within your .txt files:

Simple Choice: {boy|girl|dog}
- Randomly selects one of the items.
Weighted Choice: {0.1::rare|common}
- The value before :: is the probability weight (Default: 1.0). In this example, "rare" is much less likely to be chosen.
Dynamic Negative Extraction: -monochrome or -(bad quality:1.2)
- Any word starting with a hyphen - is automatically moved to the Negative Prompt output, and the hyphen is removed.

Outputs:

positive_prompt: Combined prompt after processing logic.
negative_prompt: Extracted negative keywords (from hyphen - syntax).

Click to see Advanced Syntax & File Rules

File & Pattern Rules

Directory: Specify the folder where your .txt files are stored.
File Pattern: Filters files (e.g., *.txt). Files starting with _ are ignored.
Selection Method:
- random: Picks a random line from each file.
- sequential: Picks a line based on the seed.
Use Break: If enabled, joins prompts from different files using the BREAK keyword.

Commenting System

You can keep your prompt files organized with standard comment syntax:

// comment or # comment: Single-line comments.
/* comment */: Multi-line block comments.

Node Usage: Meld Auto-Exposure

Provides professional-grade luminance correction to stabilize your generations.

Analyze & Adjust: Automatically analyzes image luminance and adjusts brightness/contrast to optimal levels.
Production Ready: Perfect for pre-processing images for ControlNet or img2img pipelines where lighting consistency is key.

Installation

Option 1: Comfy Registry (Recommended)

comfy node install nodemeld/comfyui-meld

Option 2: Manual Clone

cd custom_nodes
git clone https://github.com/NodeMeld/ComfyUI-Meld.git

Security & Privacy

We prioritize system safety:

Scoped File Access: Reads files exclusively from the specified directory and its subdirectories. Please ensure directory permissions are managed appropriately at the system level.
Selective Metadata Extraction: When extracting data from images, we focus only on generation-related tags (workflow, prompt, etc.). Private EXIF data like GPS coordinates are naturally ignored as they are not part of our extraction logic.

License

This project is licensed under the Apache License 2.0. It includes explicit patent grants, making it safe for both individual and enterprise use.

Author: NodeMeld