🖼️📊 Metadata Extractor (Single)

This is the node for "I have a PNG and I want to know exactly how it was made." Point it at a single image and it reads back the positive prompt, the negative prompt, and every generation parameter it can find - both as parsed values and as the raw metadata blob, in case the parsing missed something you need.

How it works

It leans on the same convention that makes ComfyUI workflow sharing possible in the first place: generation parameters get embedded directly inside the image file, not stored separately. This node reads that back out. filter_params lets you narrow the result down to specific keys - a comma-separated list like steps,sampler,seed - instead of digging through the full JSON yourself every time you just need one or two values.

The inputs and outputs that matter

• seed - selects which file to load when reading from a directory; set "increment" to cycle through one at a time, same pattern as this pack's Load Images From Path.
• input_path (optional) - a single image file or a directory of images.
• image_input (optional) - a connected image, takes priority over input_path when present.
• filter_params (optional) - a comma-separated list of the specific keys you want pulled out (e.g., steps, sampler, seed).

Outputs: image, positive_prompt, negative_prompt, parsed_params_json (the structured result), filtered_params_list (just the keys you asked for via filter_params), and raw_metadata_json (everything, unparsed).

How to install it

Via ComfyUI Manager: search ComfyUI-mnemic-nodes, install, restart. Manually:

cd ComfyUI/custom_nodes
git clone https://github.com/MNeMoNiCuZ/ComfyUI-mnemic-nodes

Restart ComfyUI. No model downloads or extra dependencies - pure metadata reading.

Common issues & troubleshooting

Everything comes back empty. No embedded metadata, no output - that's expected, not a malfunction. A re-exported, re-compressed, or screenshotted image has generally lost its embedded generation data entirely, and there's nothing to extract if it was never there.

positive_prompt/negative_prompt are blank but raw_metadata_json has data. Different tools write their generation metadata in different shapes, and this node's parsing targets the common conventions (ComfyUI's own format, A1111-style parameter strings). An unusual or less common generator's metadata might land entirely in raw_metadata_json without cleanly mapping into the named fields - check the raw JSON by hand in that case.

Real uses worth knowing about: recovering a prompt from an old render you forgot to document, building a dataset of prompt/settings pairs out of a folder of past generations by sweeping through with seed on increment, or just satisfying curiosity about how a shared image was actually made. It's the read side of the same convention this pack's Save Image With Metadata node writes.