EmAySee_ Mask Guard to Image

The "make sure we always have an image" shim

EmAySee_MaskGuard converts a MASK into an IMAGE - and if anything goes wrong, it hands you a solid black image of a size you specify instead of erroring out. That fallback is the entire point of the "Guard" in its name. You feed it a mask and it guarantees an image tensor comes out, no exceptions, which makes it the safety net for workflows where a downstream node demands an IMAGE but the mask-producing branch can come up empty (say, a detector that found nothing).

Where this matters: lots of ComfyUI consumers - previews, image compositors, some VAE-adjacent tooling - want an IMAGE and will choke or silently misbehave if handed a bare mask. If you're building a template workflow for people to reuse, wrapping the mask path in a guard node turns "crashes when the detection comes up empty" into "renders black and keeps going." It's defensive plumbing, and defensive plumbing has a real job in a graph language where one None can kill a run.

How it works

In the success path it takes the first channel of the mask, reshapes it, and replicates it across three channels so it becomes a valid RGB image tensor (white mask → white pixels, black → black). That's a literal mask-to-grayscale-image conversion. The guard part: if the mask is None or empty, or the conversion throws for any reason, it returns a 1 × height × width × 3 tensor of zeros - a black image sized by the width and height inputs.

That last bit is the subtle part: width and height only matter in the failure path. In the success path the output takes the mask's own shape, and the width/height inputs are ignored. So they're your "fallback canvas size," not a resize control.

Inputs and output

• mask (MASK) - the mask to convert.
• width (INT, 1–8192, default 512) - fallback width when the mask is unusable.
• height (INT, 1–8192, default 512) - fallback height.

Output: IMAGE - the converted mask, or the black fallback.

Install

From ComfyUI_EmAySee_CustomNodes:

cd ComfyUI/custom_nodes
git clone https://github.com/EmAySee/ComfyUI_EmAySee_CustomNodes
# restart ComfyUI

Or via ComfyUI Manager → "ComfyUI_EmAySee_CustomNodes". No dependencies.

Gotchas

The big one is the silent black image. Because a failed mask returns a black frame instead of an error, a detector that quietly finds nothing downstream looks like "the workflow ran fine, output is black." That's by design - the guard trades crash-diagnostics for continuity - so if you're debugging a black result, check whether a mask upstream actually produced content. Also, this is not an image-to-image tool: the mask's values map directly to brightness, so a soft-edged mask comes out as a soft-edged gray image, which is what you'd expect but worth remembering when the "image" looks washed out. The display name on the node even has a stray underscore ("EmAySee_ Mask Guard to Image") - cosmetic, but it's the kind of polish detail that tells you this pack is built by one person for their own graph.