AIHub Expose Image (Info Only)

AIHub Expose Image (Info Only) is the sibling of AIHubExposeImage with the pixels stripped out. Same type options, same canvas-layer semantics - it just doesn't hand you the IMAGE or MASK. What you get back is geometry only: POS_X, POS_Y, LAYER_ID, WIDTH, HEIGHT.

Why would you want that? Because sometimes you don't need to process the image contents - you need to know where it is so you can place something else there. Say your workflow generates a new texture or a background and you want to composite it back into the editor exactly where the current layer sits. One Info Only node tells you the layer's canvas position and size; the AIHub action nodes that write results back (like AIHubActionNewLayer) can use those coordinates to drop the new layer in the right spot. It's also handy as a cheap "is this layer even present" check: if optional is on, a missing image just yields zeros rather than aborting the run.

What you set

The inputs are the same as the full image expose, so the pattern is familiar:

• type - the image source, defaulting to upload. The seven options mirror the full node: current_layer, merged_image, current_layer_at_image_intersection, the *_without_current_layer variants, and upload. Since you're only getting placement data, current_layer or merged_image is the pragmatic choice - the intersection variants matter when you need to reason about exactly what region a flattened canvas covers.
• id / label / tooltip - the protocol identity and the human-facing name, same as every expose.
• optional - with it true, a missing image won't raise; you'll get zeros instead of geometry.

That's honestly the whole node. It's one of the thinnest in the pack, and it's not trying to be more.

Install & gotchas

Same story as the rest of the pack - ComfyUI Manager → search ComfyUI-aihub-workflow-exposer, or:

cd ComfyUI/custom_nodes
git clone https://github.com/otavanopisto/ComfyUI-aihub-workflow-exposer

then restart ComfyUI. No extra Python dependencies (no requirements.txt), no model files. Remember the AIHub folders (ComfyUI/aihub/...) and an AIHubWorkflowController with a unique id.

The one real gotcha is the same as its sibling: in the bare web UI this node has nothing to expose. The data only arrives when an AIHub client connects over the websocket (port 8111) and runs the workflow with a value for this expose. If you need both pixels and position, use the full AIHubExposeImage - the info-only variant is purely the cheap geometry read, and mixing them up is a classic way to end up with a workflow that silently has no image input wired anywhere.