Save Mask Bridge Simple (Non Output) (Yogurt Nodes)

Save Mask Bridge Simple (Non Output) is the most minimal save node in the Yogurt mask family, and it's the one I'd reach for when a mask is strictly bookkeeping: it saves the file, passes the mask through, and never announces itself as a workflow result.

The inputs are exactly four. masks in, and then output_dir (blank → the ComfyUI output folder), filename_prefix (supports %date:yyyy-MM-dd% tokens and node-value placeholders), and overwrite (off by default, so filenames auto-increment). It writes grayscale PNG, embeds the workflow metadata, and outputs masks, width, height, batch and saved_paths.

The "Non Output" part is what makes this specific node useful. ComfyUI gathers output nodes' results into the API response and the frontend history. A mask you're saving just to keep a record of the selection region shouldn't be in either. This node keeps the artifact on disk and out of the results stream - which matters if you're driving ComfyUI from code, or if the save lives inside a reusable subworkflow where you don't want bookkeeping bubbles floating up to the parent.

Install

All Yogurt nodes come from one pack, one install:

cd ComfyUI/custom_nodes
git clone https://github.com/yogurt7771/ComfyUI-YogurtNodes.git
cd ComfyUI-YogurtNodes
pip install -r requirements.txt

Or ComfyUI Manager → ComfyUI-YogurtNodes → install → restart. It's under Yogurt Nodes → IO.

Gotchas

The two that catch everyone in this family. First, non-output means no entry in the frontend gallery - that's the feature, not a failure; check the output folder or the saved_paths output to confirm the write. Second, overwrite defaults to false, so if you're saving to a fixed path that another tool polls, flip it to true or you'll be accumulating numbered duplicates. And since this variant gives you no format control, remember you're always getting PNG - which for masks is the right default anyway, because lossy formats blur the selection edge and shift which pixels survive your threshold on reload.