Save Folder

Small node, small job. It lets you choose where Searge writes your generated images - ComfyUI's output folder or its input folder - and passes that choice along as a typed value. Nothing else happens here.

It's part of the older Searge building-block set (Searge/_deprecated_/Inputs), from the v2/v3 era when each setting got its own little node and they were wired together into the save logic. The current v4.x workflow folds this decision into its bigger settings nodes, but this one still resolves and works if you're on an older graph.

How it works

You pick from a two-item dropdown, and the node emits the choice as the SAVE_FOLDER type. A downstream Searge save node reads that value and drops the finished image into the matching directory. The reason "input folder" is even an option: if you save a generation into input, it's immediately available as a source image for the next img2img or inpainting pass without you shuffling files around. That's the one genuinely useful trick this node enables - a quick loop where today's output becomes tomorrow's input.

The inputs and outputs that matter

This is as short as node docs get:

• save_to - an enum with two values: output folder (the default, where finished images normally go) or input folder (handy when you want to immediately reuse the result as a source image).

The output is the single save_to value, typed as SAVE_FOLDER for the save node to consume.

How to install it

Grab the whole pack; there's no single-node install.

ComfyUI Manager: search SeargeSDXL, install, restart. Manual: python -m pip install opencv-python once in ComfyUI's Python env (the pack requires OpenCV and won't fully load without it), then cd ComfyUI/custom_nodes && git clone https://github.com/SeargeDP/SeargeSDXL.git, and restart. Windows portable users can run the pack's installer script instead, which also offers to download the SDXL checkpoints and upscalers the main workflow needs.

This node needs no model files of its own.

Where people get burned

• Losing track of where images went. If you set this to input folder and then wonder why your output directory is empty, that's this switch. It's not a bug, it's the setting.
• Overwriting your source images. Saving into input alongside your real source files can clutter or shadow them. Fine for a deliberate feedback loop, annoying if you forget it's on.
• Building something new around it. It's deprecated. In the current workflow, the save destination is handled inside the newer parameter nodes - use those unless you're keeping an old graph alive.