SorenWeile-ComfyUI_MetaSaver/QUICKSTART.md

MetaSaver Quick Start Guide

TL;DR

1. Add "Save Image with Custom Metadata" node
2. Fill in field names and connect values
3. Leave unused fields empty
4. Done!

How It Works

The node shows you pairs of inputs like this:

meta_name_0: [text input]    meta_value_0: [any connection]
meta_name_1: [text input]    meta_value_1: [any connection]
meta_name_2: [text input]    meta_value_2: [any connection]
...

To Add Metadata:

1. Type a name in meta_name_X (e.g., "seed", "prompt", "cfg_scale")
2. Connect a value to meta_value_X (from any node output)
3. Repeat for as many fields as you need
4. Leave the rest empty

Example:

meta_name_0: "seed"              meta_value_0: ← connected from KSampler seed
meta_name_1: "positive_prompt"   meta_value_1: ← connected from CLIP Text Encode
meta_name_2: "cfg_scale"         meta_value_2: ← connected from KSampler cfg
meta_name_3: ""                  meta_value_3: [empty - ignored]
meta_name_4: ""                  meta_value_4: [empty - ignored]

What Gets Saved?

Your metadata is embedded in the PNG file in two formats:

1. Structured JSON (in custom_metadata field):

   {
     "seed": 12345,
     "positive_prompt": "beautiful landscape",
     "cfg_scale": 7.5
   }
   

2. Individual fields (as meta_seed, meta_positive_prompt, etc.)

Reading Metadata

Drag back into ComfyUI

Just drag the PNG into ComfyUI to restore the workflow!

Python

from PIL import Image
import json

img = Image.open("output.png")
metadata = json.loads(img.info["custom_metadata"])
print(metadata)

Command Line

python examples/read_metadata.py output.png

Tips

• Empty = Ignored: Fields with empty names are skipped
• Any Type: meta_value_X accepts ANY data type
• Standard vs Dynamic:
  • Standard = 10 field pairs
  • Dynamic = 20 field pairs
• No Limit on Name Length: Field names can be descriptive

Common Patterns

Pattern: Track Everything

meta_name_0: "seed"         → KSampler seed
meta_name_1: "steps"        → KSampler steps
meta_name_2: "cfg"          → KSampler cfg
meta_name_3: "sampler"      → KSampler sampler_name
meta_name_4: "scheduler"    → KSampler scheduler
meta_name_5: "model"        → CheckpointLoader name

Pattern: Just Prompts

meta_name_0: "positive"     → Positive CLIP Text
meta_name_1: "negative"     → Negative CLIP Text

Pattern: A/B Testing

meta_name_0: "test_id"      → Custom string "test_001"
meta_name_1: "variation"    → Custom string "red_background"
meta_name_2: "seed"         → KSampler seed

Troubleshooting

Q: I don't see all the fields! A: Scroll down in the node - they're all there (10 or 20 pairs depending on version)

Q: My metadata isn't saving! A: Make sure the field name is filled in (not empty)

Q: Can I use the same field name twice? A: Yes, but the last one will overwrite earlier ones. Use unique names.

Q: What if I need more than 10/20 fields? A: Use two MetaSaver nodes, or open a GitHub issue to request more!

Next Steps

• See USAGE.md for detailed examples
• See README.md for full documentation
• Use examples/read_metadata.py to inspect saved metadata

Happy saving! 🎨