Load Image From Path With Meta

The most useful metadata is the kind you didn't save. JDC_ImageLoaderMeta ("Load Image From Path With Meta") is the same path-or-URL loader as JDC_ImageLoader, but it also reads the generation data buried in the image's PNG metadata and hands it to you as usable outputs: the prompt, the negative prompt, and the dimensions. It's the "so that's what made this image" node.

It's part of comfy-plasma by Jordach, and it's quietly one of the more impressive pieces in the pack, because the metadata parsing is genuinely broad.

What it returns

Six outputs: image, mask, PROMPT, NEGATIVE, WIDTH, HEIGHT. Same image/mask behavior as the plain loader - path with quotes stripped, URL support, EXIF rotation fixed, alpha-derived mask. The extras are where the value is: wire PROMPT and NEGATIVE into your CLIPTextEncode nodes and you can instantly reconstruct someone else's generation from a single PNG.

The metadata parsing (this is the good part)

The loader understands most of the PNG metadata formats that were floating around:

• A1111 / Stable Diffusion WebUI - the classic parameters blob with positive, Negative prompt:, and Steps: markers.
• Easy Diffusion - both its JSON prompt styles.
• InvokeAI - both the modern sd-metadata format and the legacy Dream format.
• NovelAI - reads its Software tag and pulls the prompt from the description.
• Draw Things (iOS/macOS) - parses the XMP user comment.

If the format isn't recognized, it degrades gracefully: PROMPT and NEGATIVE come back as empty strings, and you still get the image, mask, and dimensions. So it never breaks on unknown metadata - you just don't get the prompt. And WIDTH / HEIGHT are read from the actual image, so those are always reliable.

When you'd reach for it

• You downloaded (or were sent) an image and want to recreate it - load it, grab the prompt, feed it through a sampler.
• Studying other people's workflows: same image, same seed, same prompt, tweak the one variable.
• Batch forensics: point it at a folder of PNGs and chain the outputs to see what each one was built from.

Install

No requirements.txt, no models. ComfyUI Manager → Install Custom Nodes → search comfy-plasma → install, restart. Or:

cd ComfyUI/custom_nodes
git clone https://github.com/Jordach/comfy-plasma

Restart ComfyUI. It's under image.

Common issues

• Empty prompt strings are normal for JPEGs (no PNG metadata), for images saved without embedded workflow data, or for formats the parser doesn't know. Check whether the source app actually embeds parameters before assuming it's broken.
• URL loading + metadata only works when the remote server serves the full file; some image hosts strip metadata on upload. If the prompt comes back empty from a URL, download the file and load it locally.
• Like every node in this pack, it's single-image: one file in, one full metadata readout out.