A ComfyUI custom node that extracts workflow metadata from PNG images and video files (MP4, AVI, MOV, MKV, WebM) generated by ComfyUI. Recover your prompts, settings, and generation parameters from any ComfyUI output!

📋 Features

✅ Dual format support - Extracts from both PNG images and video files
✅ Complete metadata extraction - Prompts, seed, steps, CFG, sampler, and more
✅ Multiple input methods - Dropdown selector, manual path, or tensor input
✅ Video format support - MP4, AVI, MOV, MKV, WebM
✅ Full JSON output - Access complete workflow data
✅ Video properties - Resolution, FPS, duration, codec info
✅ Easy to use - No complex configuration needed

🎯 What It Extracts

From Images (PNG)

Positive prompt
Negative prompt
Seed
Steps
CFG scale
Sampler name
Full workflow JSON
All node parameters

From Videos (MP4, etc.)

All image metadata fields
Video resolution (width x height)
Frame rate (FPS)
Duration
Video codec
Complete workflow data

🚀 Installation

Method 1: ComfyUI Manager (Recommended)

Open ComfyUI Manager
Search for "Metadata Extractor"
Click Install
Restart ComfyUI

Method 2: Manual Installation

Navigate to your ComfyUI custom nodes directory:
```
cd ComfyUI/custom_nodes/
```

Clone this repository:

git clone https://github.com/ai-joe-git/ComfyUI-Metadata-Extractor.git

Install ffmpeg (required for video support):

Windows:

winget install ffmpeg

macOS:

brew install ffmpeg

Linux:

sudo apt install ffmpeg  # Debian/Ubuntu
sudo dnf install ffmpeg  # Fedora

Restart ComfyUI
The node will appear as "📹 Extract Metadata (PNG/Video)" in the node menu under utils/metadata

💡 Usage

Basic Setup

Add the node to your workflow
- Search for "Extract Metadata" or "📹 Extract Metadata (PNG/Video)"
Choose input method (pick one):

Option A: Dropdown (Easiest)
- Copy your image/video to ComfyUI/input/ folder
- Select from the filename dropdown
- No typing needed!
Option B: Manual Path
- Paste full path in file_path field
- Example: C:\Users\YourName\video.mp4
Option C: Video Tensor (Limited)
- Connect Load Video output to video input
- Node will attempt to find source file
- ⚠️ Not as reliable - use dropdown instead
Connect outputs to Show Text nodes or use in your workflow

Example Workflow

[Extract Metadata Node]
    ├─ positive_prompt → [Show Text / CLIPTextEncode]
    ├─ negative_prompt → [Show Text]
    ├─ seed → [Show Text / Display Integer]
    ├─ steps → [Show Text / Display Integer]
    ├─ cfg → [Show Text / Display Float]
    ├─ sampler → [Show Text]
    ├─ metadata_json → [Show Text / Save to File]
    └─ file_info → [Show Text]

📤 Outputs

The node provides 9 outputs:

Output	Type	Description
`positive_prompt`	STRING	Your main generation prompt
`negative_prompt`	STRING	Your negative prompt
`seed`	INT	Random seed used
`steps`	INT	Number of sampling steps
`cfg`	FLOAT	CFG scale value
`sampler`	STRING	Sampler name (e.g., "euler_ancestral")
`scheduler`	STRING	Scheduler name (if available)
`metadata_json`	STRING	Complete workflow as JSON
`file_info`	STRING	Video/image properties summary

🎨 Use Cases

Recreate Past Generations

Extract settings from an old image to reproduce the same results:

Load old image → Extract Metadata → Use settings in new workflow

Organize Your Outputs

Extract and log all generation parameters for your archive:

Extract Metadata → metadata_json → Save to database/spreadsheet

Study Successful Prompts

Analyze what worked well in previous generations:

Extract prompts from best outputs → Study patterns → Refine workflow

Batch Analysis

Process multiple files to understand your generation patterns:

For each output:
  Extract Metadata → Aggregate statistics → Identify trends

🔍 Console Output

When extracting metadata, you'll see helpful debug info:

[Metadata] 📂 Processing: video.mp4
[Metadata] Video tags: ['comment', 'encoder']
[Metadata] ✅ Found 'comment' tag
[Metadata] Parsing 45 nodes...
[Metadata] ✅ Positive prompt
[Metadata] ✅ Negative prompt  
[Metadata] ✅ Seed: 123456
[Metadata] ✅ Steps: 20
[Metadata] ✅ CFG: 7.5
[Metadata] ✅ Sampler: euler_ancestral

🔧 Requirements

ComfyUI (latest version recommended)
ffmpeg (required for video support)
Python 3.8+ (included with ComfyUI)
PIL/Pillow (included with ComfyUI)

⚙️ Technical Details

Supported Video Formats

MP4 (H.264, H.265)
AVI
MOV
MKV
WebM

Metadata Sources

The node searches for ComfyUI metadata in:

PNG: workflow and prompt chunks
Video: comment, description, workflow tags

Node Detection

Automatically identifies and extracts from:

CLIPTextEncode - Prompts
RandomNoise - Seed values
KSamplerSelect - Sampler names
LTXVScheduler, BasicScheduler - Steps
CFGGuider - CFG scale
And many more ComfyUI nodes!

💡 Tips & Tricks

Use the dropdown - Most reliable method for selecting files
Copy to input folder - Makes files easily accessible via dropdown
Check console - Detailed logging shows exactly what's being extracted
Full JSON available - metadata_json output contains everything
Video tensor limitations - Direct tensor input can't reliably find source files
No metadata? - File must be generated with "Save Metadata" enabled

⚠️ Troubleshooting

"ffprobe not found" Error

Solution: Install ffmpeg (see installation section)
Verify: Run ffprobe -version in terminal

"No metadata found"

Check: File was generated by ComfyUI with metadata enabled
Try: Use workflow JSON export instead of direct file
Note: Not all generators save metadata

Empty Outputs

Reason: Metadata structure not recognized
Solution: Check metadata_json output for raw data
Workaround: Parse JSON manually if needed

Video Tensor Not Working

Expected: Video tensors don't contain filename info
Solution: Use dropdown or manual path instead
Why: ComfyUI design limitation, not a bug

🤝 Contributing

Contributions are welcome! Areas for improvement:

Support for more metadata formats
Additional output formats (CSV, database, etc.)
GUI for batch processing
Integration with other ComfyUI tools
Better video tensor source detection

📄 License

MIT License - Free to use in your projects!

🔗 Links

🙏 Acknowledgments

Built for the ComfyUI community to make metadata extraction simple and accessible.

Special thanks to:

ComfyUI developers for the amazing framework
FFmpeg project for video processing capabilities
Community members who suggested features and improvements

Made with ❤️ for the ComfyUI community

If this node helps your workflow, please star the repository! ⭐

📸 Screenshots

Node Interface

The node provides three input options for maximum flexibility:

filename: Dropdown selector for files in input folder
file_path: Manual path input
video: Direct tensor connection (limited)

Output Example

positive_prompt: "a beautiful landscape with mountains and rivers, sunset lighting"
negative_prompt: "blurry, low quality, distorted"
seed: 424242
steps: 20
cfg: 7.5
sampler: "euler_ancestral"
file_info: "1920x1080 | 24.00 FPS | 5.00s | h264"

Need help? Open an issue on GitHub or reach out to the ComfyUI community!