Up to 10 Metadata Fields: Standard version with 10 optional metadata slots
Up to 20 Metadata Fields: Dynamic version with 20 optional metadata slots
Custom Field Names: Give each metadata field a custom name
Flexible Input Types: Supports ANY type - strings, integers (seeds), floats (CFG), and more
PNG & Video Metadata Embedding: Custom data saved in PNG files or video container tags
Workflow Preservation: Maintains standard ComfyUI workflow metadata
Multiple Formats: Metadata saved both as structured JSON and individual fields
All Optional: Only fill in the fields you need - empty fields are ignored
Video Output: Save MP4 (H.264) or WEBM (VP9) video from image frame batches

Why MetaSaver?

Unlike standard save nodes, MetaSaver lets you:

Save seeds from KSampler nodes
Store positive and negative prompts
Record model names, LoRA weights, CFG values
Add any custom text or numbers you want to track
Name your fields exactly how you want them
Embed metadata in videos as container-level tags

Perfect for tracking generation parameters, comparing outputs, or sharing images and videos with full context!

Installation

Method 1: ComfyUI Manager (Recommended)

Open ComfyUI Manager
Search for "MetaSaver"
Click Install

Method 2: Manual Installation

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

Clone this repository:

git clone https://github.com/SorenWeile/ComfyUI_MetaSaver.git

Restart ComfyUI

Usage

Saving Images

Add the "Save Image with Custom Metadata" node to your workflow
You'll see pairs of inputs for metadata (10 pairs in standard, 20 in dynamic version):
- meta_name_0: Name of first field (e.g., "seed")
- meta_value_0: Value to save (e.g., connect from KSampler seed output)
- meta_name_1: Name of second field (e.g., "positive_prompt")
- meta_value_1: Value to save
- ... and so on
Fill in only the metadata fields you want to use (empty ones are ignored)
Connect your image output to the images input
Run your workflow!

Saving Videos

Add the "Save Video with Custom Metadata" node to your workflow
Connect an IMAGE batch (your video frames) to the images input
Set fps, choose format (mp4 or webm), and fill in any metadata fields
Run your workflow — the video is saved to your ComfyUI output folder

A typical video workflow: KSampler → VAE Decode → Save Video with Custom Metadata

Or for multi-frame video: connect the image batch output from a video model directly.

Note: All metadata fields are optional - you can use as many or as few as you need!

Reading Metadata

From PNG images:

ComfyUI: Drag the image back into ComfyUI to see all metadata

Python:

from PIL import Image
img = Image.open("your_image.png")
metadata = img.info.get("custom_metadata")
print(metadata)

ExifTool: exiftool your_image.png

From MP4/WEBM videos:

Command line: ffprobe -v quiet -print_format json -show_format your_video.mp4

Python:

import av, json
with av.open("your_video.mp4") as container:
    custom = json.loads(container.metadata.get("custom_metadata", "{}"))
    print(custom)

ExifTool: exiftool your_video.mp4

The metadata JSON structure is the same for both images and videos:

{
  "seed": 12345,
  "positive_prompt": "a beautiful landscape",
  "cfg_scale": 7.5,
  "model_name": "sd_xl_base_1.0"
}

Node Variants

MetaSaver (Standard)

Saves a PNG image with 10 optional metadata field pairs (meta_name_0 through meta_name_9). Perfect for most use cases.

MetaSaverDynamic (Advanced)

Saves a PNG image with 20 optional metadata field pairs (meta_name_0 through meta_name_19). For power users who need to track many parameters at once.

MetaVideoSaver (Standard)

Saves a video (MP4 or WEBM) from an IMAGE batch with 10 optional metadata field pairs. Metadata is embedded as container-level tags readable by ffprobe, ExifTool, or PyAV.

MetaVideoSaverDynamic (Advanced)

Saves a video (MP4 or WEBM) from an IMAGE batch with 20 optional metadata field pairs.

Common Use Cases

Track Generation Parameters:
- Seed, steps, CFG scale, sampler name
- Easy to reproduce exact results later
Model Comparison:
- Save model name, version, hash
- Compare outputs from different models
Prompt Engineering:
- Store positive/negative prompts with images
- Build a library of what works
LoRA Experimentation:
- Record LoRA names and weights
- Track which combinations work best
Batch Processing:
- Add batch IDs, timestamps, custom notes
- Organize large generation runs

Technical Details

Image Format: PNG with embedded metadata (PIL/Pillow)
Video Formats: MP4 (H.264, yuv420p) or WEBM (VP9, yuv420p) via PyAV
Metadata Storage: Structured JSON + individual text fields (images); container-level tags (videos)
MP4 Metadata: Written using movflags=use_metadata_tags for full tag support
ComfyUI Integration: Follows standard node conventions
Workflow Compatibility: Preserves standard ComfyUI workflow data (prompt + extra_pnginfo)

Troubleshooting

Node doesn't appear in ComfyUI

Restart ComfyUI completely
Check that the folder is in ComfyUI/custom_nodes/
Look for errors in the console

Metadata not saving

Ensure field names are not empty
Check that values are connected to inputs
For images: verify PNG format is selected
For videos: use ffprobe -show_format_tags to inspect saved file

Can't read metadata

Images: use exiftool or Python PIL — check the custom_metadata field for structured JSON
Videos: use ffprobe -v quiet -print_format json -show_format your_video.mp4
Individual fields are prefixed with meta_ in the Dynamic variant

Video node not available

Ensure PyAV is installed: pip install av
PyAV is bundled with most ComfyUI distributions — check the console for import errors

Contributing

Contributions are welcome! Please:

Fork the repository
Create a feature branch
Submit a pull request

License

MIT License - see LICENSE file for details

Credits

Inspired by excellent existing nodes:

Support

Found a bug or have a feature request?

Open an issue on GitHub
Provide workflow examples if possible
Include ComfyUI console errors

Enjoy tracking your AI generations with full metadata control! 🎨✨