Clean Save Image

ComfyUI has a culture where the workflow is the artifact: the stock SaveImage node embeds the entire graph, prompt, and settings into the PNG metadata, and "workflow included" is a community norm. That's great when you're sharing a workflow with someone who'll drag the image back into ComfyUI and reconstruct it. It's a liability when you're selling a print, posting to a platform that strips or reads EXIF, or just don't want your entire pipeline shipped inside every file you upload.

Clean Save Image is the fix: a drop-in replacement for SaveImage that writes the pixels and nothing else. No EXIF, no PNG tEXt/iTXt/zTXt chunks, no embedded workflow, no AI-generation comments. The source is blunt about it - the node receives the prompt and extra_pnginfo in its hidden inputs and deliberately discards them. "That is the point," says the code comment.

The inputs that matter

• images - the image batch to save, straight from a VAE decode or an upscale.
• filename_prefix - prefix for the saved files, default ComfyUI, written to the standard ComfyUI output folder with the usual name_00001_.ext counter pattern.
• format - png, jpeg, or webp. The stock node is PNG-only, so this alone is a reason to reach for it.
• quality - 1–100, default 90, used for JPEG/WEBP (ignored for PNG).
• lossless_webp - WEBP only: flips on lossless encoding.

No outputs - it's an output node, it terminates the graph, which is exactly where a save belongs.

The trade-off nobody tells you about

Stripping metadata cuts both ways, and it's worth being deliberate. The reason people can drag a PNG back into ComfyUI and get a working graph is precisely the metadata this node deletes. Once you save with Clean Save Image, that file is a dead end - the image is yours, but the recipe is gone. So the sensible split is: use this for anything leaving your machine (sales, galleries, profiles), and keep a stock SaveImage in the same workflow when you want a self-documenting copy for your own archive. Two save nodes, two purposes, no conflict.

Installing it

It's part of MoonPack:

cd ComfyUI/custom_nodes
git clone https://github.com/moonwhaler/comfyui-moonpack.git

or ComfyUI Manager → search MoonPack → install, then restart. Under MoonPack/image. No models, no extra Python dependencies.

Bottom line

A privacy scrubber and a format chooser in one tiny node, and it composes with the rest of the pack's "output what you want, not what the framework wants" ethos. If you've ever shipped a PNG that leaked your whole workflow, you'll get why this exists the first time you use it.