Get Values from Image

Every query-based tool has the same problem: you need to know what to ask for. Get Values from Image is the pack's answer - a node that doesn't take a query at all. It dumps every recoverable value from an image's metadata as a list of #NODE_ID.WIDGET_NAME: value lines, aligned into columns, so you can see exactly what's available and copy a key into one of the pack's other nodes.

It's part of shinich39's comfyui-get-meta, and the pack itself labels it "for debugging" in its display names. That's not an insult - it's the job description. When you don't know the node ID or exact widget name for the seed, denoise, or prompt you want, drop this in first, read the list, then swap it for the targeted node.

How it works

The frontend extension reads the image metadata the same way every node in the pack does - trace the loader feeding image, have the backend open the file with Pillow, parse the embedded prompt JSON. Then instead of resolving one query, it walks every node's inputs and builds a line per widget, like:

#3.prompt              : a photograph of a lighthouse at dusk, ...
#4.ckpt_name           : dreamshaper_8.safetensors
#7.seed                : 123456789
#7.cfg                 : 7
#7.denoise             : 0.6

String values get whitespace collapsed, nested object values (like conditioning) are skipped because they're not extractable, and lines are padded so the columns line up. The Python node is a passthrough that returns the widget text - the always-true IS_CHANGED ensures it re-runs with fresh data each queue.

The inputs and output that matter

Just two inputs, both required:

• image - from a supported loader (core LoadImage/LoadImageMask, Inspire Pack, Crystools, WAS Node Suite).
• nodes - the multiline widget that gets filled with the listing. Read it, don't edit it.

The single STRING output carries the full listing, so you can pipe it to a text-display node or save it if you need the whole inventory somewhere persistent.

Installing it

ComfyUI Manager → search comfyui-get-meta → Install, restart. Or:

cd ComfyUI/custom_nodes
git clone https://github.com/shinich39/comfyui-get-meta

Then restart. No models, no extra dependencies beyond what ComfyUI already ships.

Common issues

• Empty list - the image has no embedded metadata. Stripped saves, web re-uploads, screenshots: nothing to enumerate.
• The listing looks like gibberish - that's expected. You're reading a dump of widget values, not a human summary. The point is the keys: note the #7.seed format and use that exact string in another Get node.
• Unsupported loader - video-frame and URL loaders aren't traced, so the widget never fills.

Honest take: this is a debugger, and it knows it. It's rarely in your final workflow - you use it once to discover the keys, then replace it with the typed Get node that fits. But when you're reverse-engineering someone else's image and have no idea what's inside, it turns guesswork into a menu.