Simple Readable Metadata-SG

Every ComfyUI image carries its whole recipe inside the file - the prompt, LoRA, model, seed, sampler, scheduler, all of it. That's the famous "workflow included" culture: drag an image back into ComfyUI and the graph rebuilds. The catch is that the recipe is stored as a JSON blob that's basically unreadable to human eyes. Simple Readable Metadata-SG fixes exactly that: you drag in an image and it prints the generation settings as plain, formatted text you can actually skim.

The pack's flagship node is SimpleReadableMetadataSG (category image/analysis). It loads a PNG or WEBP from your ComfyUI input folder (or drop one in), reads the embedded metadata, and hands you a readable summary - prompt, negative prompt, model, LoRA, seed, steps, CFG, sampler, scheduler - plus typed outputs for the individual values.

What makes it different from dragging the image back in

Drag-and-drop reconstruction works only for images saved by ComfyUI itself, and only when the full workflow was embedded. This node also understands the format used by Automatic 1111 and Forge: those save their generation settings as a single parameters string inside the PNG info chunk, and there's no graph to drag back at all. If you've ever downloaded an image from Civitai and wondered what seed or sampler made it, this is the node. The source checks for ComfyUI's prompt/workflow keys and the A1111 parameters key, then parses whichever one it finds into the same readable layout.

The inputs that matter

Only three, and you'll mostly touch the first one:

• image - pick from the dropdown (it lists files already in ComfyUI/input) or drag a file straight onto the widget.
• emoji_in_readable_text (default on) - prefixes the summary with 🎨📦 style labels. Harmless flavor; turn it off if you're saving the output to a file or piping it somewhere that hates emoji.
• show_info (default both) - choose properties (resolution, filename, file size), metadata (the generation settings), both, or none.

The outputs worth wiring

The big one is Simple_Readable_Metadata (STRING) - the whole formatted summary. Feed it into the pack's Text Viewer or Save Text node and you've got a permanent record. But the typed outputs are the real point if you want to act on the data: Positive_Prompt, Negative_Prompt, seed (INT), metadata_raw (the untouched JSON/parameters string), file_name_text, plus image and mask pass-throughs so the node can sit inline in an img2img pipeline without breaking it. The author built it explicitly for img2img/img2vid workflows.

Installing it

One-time pack setup, same for all seven nodes in it:

cd ComfyUI/custom_nodes
git clone https://github.com/ShammiG/ComfyUI-Simple_Readable_Metadata-SG.git
cd ComfyUI-Simple_Readable_Metadata-SG
pip install -r requirements.txt   # opencv-python, that's it

Or skip all that: open Manager → Custom Nodes Manager, search Simple Readable Metadata-SG (or the author, ShammiG), install, restart. No model downloads, no weights, nothing to place in models/. If Manager doesn't show it, refresh its node cache and search again.

Where people get burned

• The Node 2.0 warning. The README is blunt: on ComfyUI 0.3.76 or newer, flip the "Node 2.0" setting OFF or the display breaks. The ComfyUI frontend moved from Lite graph rendering to Vue, and this pack hasn't been rewritten for it yet. This is the single most likely "why is nothing showing" cause on a current install.
• Stripped metadata. Some image hosts and resizers strip PNG chunks. No metadata in the file means you'll get a "no metadata found" style message, and no tool can fix that.
• The (empty) prompt bug. If a CLIP Text Encode node's text box has a wire connected to it, ComfyUI stores the link, not the text, so the embedded workflow shows an empty prompt. That's why this pack ships Save Prompt nodes - see those pages for the fix.

Want more than a summary? SimpleReadableMetadataMAXSG is the same engine with every value as a separate typed output, and the Text Viewer adds search, copy, and export.