🐳批量删除元数据

Every ComfyUI image you save carries its whole workflow embedded in the PNG metadata. That's a feature when you want to share the recipe, and a liability when you don't - you're shipping your graph, your model paths, sometimes your prompts, to anyone who opens the file. 🐳批量删除元数据 (BatchMetadataRemover) strips that out of every image in a folder in one go. It's the "clean these before I upload them" node, and it does exactly what it says on the tin.

The mechanism is boring in the best way. It globs a folder for files matching your pattern, opens each with Pillow, and re-saves them with the metadata dropped. The source shows it re-encoding PNG at compress_level=4 and JPEG at quality 95, so files are slightly recompressed but visually identical - the whole point is to break the metadata chain. You point it at a folder, it walks the glob, and it returns a status string telling you how many succeeded and how many failed.

The inputs

• 输入文件夹 - the source folder path (plain string, no browser).
• 输出文件夹 - where cleaned files go. Default cleaned_images, relative to wherever ComfyUI runs, so give it an absolute path unless you like surprises.
• 文件匹配模式 - the glob, default *.png. Want JPEGs too? Set it to *.png then... no - use *.jpg or *.* for the whole folder. It's one pattern, not a list.
• 保留原始文件 - default on (true). Cleaned copies go to the output folder and originals stay untouched. Flip it off and it overwrites in place, which is risky - keep it on until you've verified a few files.

The single output, 处理状态, is a status STRING that tells you the count. It's marked as an output node, so the result shows in your UI.

Install and the honest use case

From comfy_Pond_Nodes:

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

No model, no heavy deps - just Pillow, which ComfyUI already has. Now the honest framing: this is a batch tool for a folder that's already on disk. If you're cleaning images as they're generated, ComfyUI's own Save Image lets you leave the workflow out - you don't need a metadata scrubber at all. Reach for this when you've got a pile of existing PNGs (downloaded workflows, a shared archive, a model's example images) and you want to strip them before re-hosting. For that job it's the right tool, and the pack's "load image (clear metadata)" sibling covers the single-file case.