CMK Image Preview Forward -Pipe-

A one-input, one-output node that does nothing to your image and still earns its place on the canvas. It hands the tensor straight through and throws a preview onto ComfyUI's normal image display, so you can eyeball what a CMK subgraph has produced without breaking the wiring to see it.

That sounds trivial until you've worked inside CMK Flow for ten minutes. CMK Image Preview Forward -Pipe- exists because of a rule the pack takes seriously: internal boundaries in CMK are mandatory, and you don't get to route a payload around a Prepare/Execute just to peek at it. A core PreviewImage in the middle of a CMK module is a wiring smell; this node is the sanctioned way to see the pixels and keep the flow intact.

What it actually does

It's a subclass of the pack's plain CMK Image Forward -Pipe-. Same body, one addition: before returning the image, it calls ComfyUI's own PreviewImage().save_images(IMAGE) and returns the resulting ui payload alongside the image.

Two things follow from that implementation:

• The preview you see is a bog-standard ComfyUI preview. Same temp files, same click-to-open, same behaviour you already know.
• The preview call is wrapped in a try/except and simply dropped on failure. The node will never fail because of the preview - a broken preview path costs you the picture on screen, not the graph. The only thing it raises on is an actually missing input: if nothing is connected, you get CMK Image Preview Forward -Pipe-: IMAGE is missing.

Inputs and outputs

| | name | type | |---|---|---| | in | IMAGE | IMAGE | | out | IMAGE | IMAGE |

That's the whole node. No mask, no batch handling, no settings to tune. It's a tap, not a filter, which is why it's filed under CMK/Developer/Pipe/Preview rather than with the image toolbox.

Install

Same as the rest of the pack - one install covers all 20-odd CMK nodes.

cd ComfyUI
git clone https://github.com/CMKFlow/cmk_nodes.git custom_nodes/cmk_nodes
python3 custom_nodes/cmk_nodes/scripts/install_cmk_requirements.py

Or ComfyUI Manager → search CMK Flow (registry name cmk-flow, publisher CMKFlow) → install → restart. If you install through the Manager the folder may be named cmk-flow; CMK doesn't care what the folder is called, but scripts then live under custom_nodes/cmk-flow/scripts/.

Your requirements.txt pull in ultralytics, onnxruntime, insightface, segment-anything, imageio-ffmpeg, nudenet==3.4.2 and opencv-python - heavy, because the pack covers detection and face work too. A manual git clone never runs pip for you, so the installer script is not optional.

The pack targets ComfyUI 0.38.2 with frontend 1.53.6, and wants Vue Nodes / Nodes 2.0 enabled: without it the dynamic CMK nodes fall back to the old LiteGraph drawing and CMK prints a warning at startup.

Where people trip

The preview only exists if the node runs. ComfyUI executes what the output nodes pull, so a preview forward sitting on a dead-end branch may never execute at all - the image looks "missing" when actually the node was skipped. Park it on the path you're actually queuing.

Second thing: it's flagged DEV_ONLY in the source. Expect it under developer/preview categories, and don't plan a workflow around it being the thing that saves your output. If you want the image on screen and saved, put a real Save Image (or the pack's own Upscale & Save module) downstream - this node is for the middle of the run, not the end.