Simple Preview SEGS

SEGS is a great data structure and a terrible thing to look at. It's a bundle of crops and masks and metadata, and unless you're fluent in reading bounding-box coordinates, "the detector found five regions" and "the detector found fifty" feel identical until you run the detailer and it does something surprising. SimpleSyrup.SimplePreviewSEGS is the visibility fix: it draws your SEGS over the original image, lets you inspect them in an interactive grid, and passes the SEGS through untouched. It changes nothing; it just stops you from working blind.

Two inputs - image (the source image the SEGS were detected from) and segs (the regions to inspect) - and one output: segs, the original SEGS passed through without modification. That passthrough is the important architectural detail. You drop this node into the middle of a wire as an inspection station: detection → preview → detailer. The preview sees everything, the detailer still gets the untouched SEGS, and you haven't introduced a single transformation into the pipeline. It's the ComfyUI idiom of putting a preview on a wire you'd otherwise have to re-derive by hand.

The interactive part is where it earns its keep. SimpleSyrup's preview runs the SEGS over the original image - much better than staring at crops in a gallery, because you can see where each region actually sits - and offers a selectable grid for inspecting individual regions. If your workflow's detector over-detects (SAM auto-mode especially loves inventing regions), this is the node where you notice that, before you've paid a detailer to redraw twelve phantom faces. ADetailer users will recognize the itch: the whole reason the ADetailer workflow gained a side-by-side face-detailer view is that you can't fix what you can't see.

It's an output node (is_output_node in the schema), so it behaves like a Preview Image - it renders, and it's the terminal of that branch. Wire it where you want to look, and don't feel bad about having two or three of them in one workflow during debugging. When you're done tuning, delete them; the SEGS value carries on either way.

Install: the pack:

cd ComfyUI/custom_nodes
git clone https://github.com/Artificial-Sweetener/SimpleSyrup
cd SimpleSyrup && pip install -r requirements.txt

or ComfyUI Manager → search SimpleSyrup → Install → restart, with a current ComfyUI (v3 extension API). No models, no heavy dependencies - it's a viewer. If the overlay doesn't look like your image, the image input is probably not the exact source the SEGS were detected from; SEGS store crop coordinates relative to their source, so mismatched sources produce nonsense overlays and that's the first thing to check.