HBH Image Coordinate Preview

If you've used HBH_ImageCoordinatePicker, you probably noticed the point never appears on its image output. That's deliberate: drawing is this node's job, not the picker's. Feed HBH_ImageCoordinatePreview the same image and the coordinates_json string, and it draws a filled circle at exactly that spot - the visible half of the picker pair.

Mechanism, from the source: it parses coordinates_json, pulls out x, y, point_color and point_size, converts the image to PIL, then paints a circle by filling every pixel whose distance from the center is within point_size. Colors are a fixed five - red, blue, green, yellow, white. Anything missing from the JSON falls back to defaults (red, size 10). If the JSON won't parse at all, the node prints an error to the ComfyUI console and returns your original image untouched - which is easy to mistake for a silent no-op.

Inputs: image and coordinates_json. The JSON can come straight off HBH_ImageCoordinatePicker or HBH_ImageInteractivePicker, both of which emit it, or from any other node that produces a compatible object. Output: the drawn IMAGE. This isn't an output node, so you'll want to run its result into a PreviewImage (or HBH_ImagePreview from the same pack) to actually see it.

And honestly, that's the whole job. It's a QA tool: pick a point, preview it, confirm it's where you think, then send the coordinates on their way. It's also a decent way to sanity-check the JSON shape before it hits HBH_ImageOverlay, since the overlay is far fussier about format.

One detail worth knowing if you mix pickers: the InteractivePicker's coordinates_json only carries x, y, width and height - no color or size - so if you feed that straight into this node, the marker silently reverts to red at size 10.

Install is shared across the pack. ComfyUI Manager: search ComfyUI_HBH-image_overlay. Or by hand:

cd ComfyUI/custom_nodes
git clone https://github.com/wings6407/ComfyUI_HBH-image_overlay

then restart ComfyUI. No requirements.txt, no models to download - just numpy, torch and Pillow, which ComfyUI already ships with.

Same pack, same caveats: solo author, essentially no community threads about it, so there's no fix pool to draw from if something misbehaves. The closest thing to documented troubleshooting is the README's tip to pass images through an RGB conversion node (WAS Suite's Images to RGB) before chaining the pack into a larger workflow.