Mask Crop Region

This is the node that makes "crop the region, fix it at full resolution, paste it back" possible in a WAS-based workflow. It finds a region in your mask - either the dominant (largest) blob or the minority (smallest) - crops to it, and critically, hands you back crop_data: the record of where it cropped from, so you can paste the fixed region back into its exact original spot later.

That crop_data output is the important part. Cropping is easy; putting the piece back where it came from without misalignment is the annoying part, and crop_data is the receipt that makes the round-trip clean.

Why you'd reach for it

This is the WAS version of the single most important pattern in modern inpainting: don't repaint the whole frame, crop out just the region that needs work, generate it at proper resolution, and composite only that piece back. A face that's 80×80 pixels in a 1024×1024 image gets almost no model attention and comes out mangled; crop it out, and now those same 80 pixels get rendered at full generation resolution. That's the entire idea behind detail passes (Impact Pack's whole detect-crop-refine loop is built on it), and it's why the 2026 community consensus is to composite the fixed region back rather than run the whole image through another VAE cycle and degrade everything.

Mask Crop Region gives you the crop and the coordinates to reverse it. Do your work on the cropped piece, then use the location/size data to paste it back exactly.

How it works

You give it a mask. It locates the region you asked for - dominant or minority - optionally pads it (so you're not cropping right against the edge of the subject), and crops to that bounding box. Alongside the cropped mask it outputs the geometry: the crop_data blob plus the region's location and size as numbers, which the pack explicitly notes you can feed into other nodes "like Crop Image Location" to crop the matching image to the same box.

The inputs and outputs that matter

• mask - the input mask to find a region in.
• region type - dominant (largest area) or minority (smallest). Pick which blob you care about.
• padding - pixels of breathing room around the region. Add some; cropping tight to the subject makes for ugly paste-backs.

Outputs: the cropped MASK, the crop_data (your paste-back receipt), and the region's location and size as numbers to drive matching image crops.

Installing it

Part of WAS Node Suite. ComfyUI Manager: search WAS Node Suite, install, restart. Or:

cd ComfyUI/custom_nodes
git clone https://github.com/WASasquatch/was-node-suite-comfyui

install requirements.txt against your ComfyUI Python, restart.

Common issues

Paste-back is misaligned. Almost always a mismatch between the crop_data / location used to crop the image and the one used to paste it back. Use the same geometry outputs for both the crop and the paste. WAS's Image Paste Crop node consumes crop_data for exactly this.

Cropped too tight, ugly seams. Bump the padding. Too little padding means the fixed region butts right against unfixed pixels with no room to blend.

It grabbed the wrong blob. Dominant vs minority is doing what you told it. If your target isn't the largest region, switch to minority - or clean the mask first (threshold/fill) so the region you want is actually the dominant one.

All WAS nodes gone after an update. Pack-level. WAS Node Suite is retired and unmaintained; its dependency pins collide with ComfyUI updates and throw "Import Failed," removing every WAS node until you reinstall the pack's requirements against the correct ComfyUI venv.

For a fully modern take on this same round-trip, the community standard is now the dedicated Inpaint Crop-and-Stitch node pair, which never runs unmasked pixels through the VAE at all. But if you're already in a WAS graph, Mask Crop Region plus Image Paste Crop does the job with tools you already have.