saveImage

ComfyUI's built-in Save Image does something sneaky by default: it embeds the entire workflow as metadata inside the PNG. That's normally a feature - it's how "drag the image back in and the graph rebuilds" became culture. But when your ComfyUI graph is your product, that same feature is a leak. Paying customers can take any image they generated, drag it back into ComfyUI, and reconstruct exactly how you made it.

sdBxb_saveImage (display name "saveImage") is the pack's fix: it saves the pixels and none of the recipe. The README spells it out plainly - the official save node automatically stores workflow info; this one doesn't.

How it actually works

It's ComfyUI's SaveImage, copied and modified in three small ways. It strips metadata entirely (it saves with pnginfo=None), so the PNG carries nothing about your graph. It prefixes every filename with sdBxb_, so your converted-app outputs are easy to spot in the output folder. And it supports %date:yyyy%-style tokens in the filename - write sdBxb_%date:yyyy-MM-dd% and each batch lands in a dated file. Nothing else changes: it's still an output node, still writes to ComfyUI/output, still handles batches.

The inputs

Only two, and both are familiar:

• images (IMAGE) - the tensor you'd normally feed a Save Image node.
• filename_prefix (STRING, default ComfyUI) - the base name, prepended with sdBxb_ automatically.

Installing it

Part of the ComfyUI_Bxb pack. Install via ComfyUI Manager (search ComfyUI_Bxb) or:

cd ComfyUI/custom_nodes
git clone https://github.com/zhulu111/ComfyUI_Bxb

Then restart. Pip dependencies install themselves on first load; there are no model files to fetch.

Common issues and the one "gotcha" that's actually a feature

The classic confusion is dragging a saved image back into ComfyUI and getting nothing - no workflow reconstructs. That's not a bug, it's the entire point of the node. If you want shareable, workflow-embedded images, use the built-in Save Image instead. In a conversion workflow, this is the node to save with so user outputs get captured by the app - the README's own framing for when it exists. One practical note: it always writes PNG regardless of what you're saving, so don't expect format options here. It's a deliberately boring node that does one job: save the image, keep the secret.