ImageCropWithBBoxList(FaceParsing)

If FaceBBoxDetect or BBoxDetect found three faces in your image, this is the node that turns those three boxes into three separate cropped images in one step, instead of you manually pulling each box out with BBoxListItemSelect and cropping one at a time. It's the batch version of ImageCropWithBBox - same underlying operation, but it takes the whole list at once.

The inputs that matter

Two, both required:

• bbox_list (BBOX_LIST) - straight from FaceBBoxDetect or BBoxDetect, one box per detected face.
• image (IMAGE) - the source image all those boxes were detected in.

The output is a single IMAGE - but functionally this is a batch/list output covering all the crops, one per box in the list. Downstream, you'd typically loop over it (ImageListSelect to pull individual crops out for per-face processing) or run FaceParse across the batch if your graph is set up to handle a list of images at once.

Where it fits

Detect faces once with FaceBBoxDetect, crop them all here, then either process them together as a batch or split back out with ImageListSelect if each face needs individual handling (different mask settings, different smoothing amounts, whatever your workflow calls for). It's the node that makes multi-face images tractable without wiring up a separate detect-crop-parse chain per face by hand.

Installing it

Comes with the pack:

• ComfyUI Manager - search "comfyui_face_parsing", install, restart.
• Manual - cd ComfyUI/custom_nodes && git clone https://github.com/Ryuukeisyou/comfyui_face_parsing.git, restart ComfyUI.

No models needed - this is pure image slicing, downstream of whatever detector produced the box list.

Common issues

If the output list comes back shorter than you expect, or missing a face you can see, that traces back to the detection step (FaceBBoxDetect/BBoxDetect), not this node - check the threshold there first. The other thing worth knowing: because each crop can be a different size (faces aren't all the same size in the source image), some downstream nodes that expect a uniform batch shape can get fussy about a list of variously-sized crops. If you hit a batch-size mismatch error further down the graph, that's usually the cause - you may need to resize each crop to a common size (ImageResizeCalculator can help you compute consistent target dimensions) before whatever comes next.