Nodes/comfy_Pond_Nodes/🐳查看元数据
ComfyUI Node

🐳查看元数据

See everything a PNG is hiding before you load or share it

By Pondowner857·Created about a year ago·Updated 21 days ago· 45
🐳查看元数据
    • 所有元数据
    • 提示词
    • 工作流
    • 图像
    上传图像

    Here's the fact that surprises everyone eventually: a ComfyUI PNG is not a picture with metadata attached. The metadata is the workflow - the full node graph, your prompt, your seed, every checkpoint and LoRA by name and hash, all embedded in the file as JSON text chunks. The image is the thumbnail of its own source code. MetadataInspector, from 🐳 Pond Nodes (comfy_Pond_Nodes), is the node that opens the file and shows you what's actually in there.

    You want it in two situations. First: you downloaded someone's workflow PNG and want to know what you're really loading before you commit. Second: you're about to post a generated image publicly and you need to see what your own file is leaking - because the default in ComfyUI is disclosure, not privacy.

    How it works

    No model, no API, no clever ML. It opens the image with Pillow and reads the PNG text chunks via img.info - the prompt and workflow chunks that ComfyUI's SaveImage writes by default, plus anything else an editor or another tool stuffed in there. What comes out is the raw truth of the file.

    The inputs and outputs

    The single required input is 上传图像 (upload image), a file picker that lists everything in your ComfyUI/input/ directory. You pick a file already sitting there, or upload one.

    Four outputs:

    • 所有元数据 - every metadata chunk the file has, dumped as formatted JSON. The kitchen sink.
    • 提示词 (prompt) - the embedded prompt graph. This is the part you can actually send to a sampler.
    • 工作流 (workflow) - the full node graph, pretty-printed JSON. This is the "loadable" workflow.
    • 图像 - the image itself, passed through so you don't need a separate load node to keep it in your graph.

    A quick inspection workflow: drop the 提示词 and 工作流 outputs into text nodes (or a save node) to keep a copy, and preview 图像.

    Installing it

    Same pack install as always. ComfyUI Manager → search comfy_Pond_Nodes → install → restart, or:

    cd ComfyUI/custom_nodes
    git clone https://github.com/Pondowner857/comfy_Pond_Nodes
    cd comfy_Pond_Nodes
    pip install -r requirements.txt
    

    This node only needs Pillow, which ComfyUI ships anyway - no heavy deps. It lives under 🐳Pond/元数据 (metadata). If your console ever starts endlessly scrolling after installing this pack, the README points at a known conflict with comfyui_HiDream-Sampler.

    The gotcha

    The workflow only survives if the file was saved by something that writes it - a ComfyUI SaveImage. The moment a PNG gets re-encoded, screenshotted, or exported to JPEG/WebP, the graph chunk is gone (JPEG and WebP have no PNG text chunks). So if you inspect a JPG and get empty outputs, that's not the node being broken; the information was destroyed before it ever reached you. And the flip side of the same coin: when you're the one sharing, this node is the mirror image of Pond's RemoveMetadata - inspect first, and if you don't want your seed and exact model list public, scrub before posting.

    Category🐳Pond/元数据

    Inputs (1)

    NameTypeDefaultDescription
    上传图像COMBO3 options: brekkie.jpg, example.png, plantpot.png

    Outputs (4)

    NameTypeDescription
    所有元数据STRING
    提示词STRING
    工作流STRING
    图像IMAGE