FC FCCropFace

FC FCCropFace is the stripped-down crop node from ComfyUI-FaceChain: detect a face, crop around it, and hand you the bounding box - without the mask and landmark outputs the fuller crop node produces. Two inputs, two outputs, no extras. It's the "I just need a face crop and where it was" option.

It's also one of the pack's older/underscore-class names. In the current node set, FC FaceDetectCrop is the flagship crop node and does detection + crop + mask + landmarks in one go; this class is the leaner predecessor. It still resolves in older workflows and pack listings, which is why it's worth a page: you'll meet it when loading a workflow someone saved before the refactor.

Inputs and outputs

• source_image (IMAGE) - the photo to crop from.
• crop_ratio (FLOAT, default 1.0, 0–10) - how much margin around the face. 1.0 is tight to the face; higher values pull in hair, forehead, and shoulders. Same parameter semantics as FC FaceDetectCrop's crop_ratio.

Outputs:

• IMAGE - the cropped face.
• BOX - the bounding box of the cropped region in the original image's coordinates, so you can paste the crop back or use it for alignment.

Under the hood it's the same ModelScope face-detection pipeline the pack uses everywhere (damo/cv_ddsar_face-detection_iclr23-damofd), downloaded on first use from modelscope.cn.

Where it fits and how it differs

The output pair - crop + box - is exactly what a crop-and-paste-back graph needs at minimum. The missing pieces versus FC FaceDetectCrop are the MASK and KEY_POINT outputs. If your downstream step needs a mask (most compositing in this pack does - see FC ReplaceByMask), this node won't give it to you, and you'll either reach for FC FaceDetectCrop instead or add a FC CropMask-style step on top.

Honest recommendation: for a new workflow, prefer FC FaceDetectCrop. This node is a clean, minimal alternative when you genuinely want just the crop and the box and nothing else - but the flagship node does all of that plus the mask and landmarks, so the extra outputs rarely cost you anything.

Install

Same pack: ComfyUI Manager (search "ComfyUI-FaceChain") or git clone https://github.com/THtianhao/ComfyUI-FaceChain into custom_nodes/, restart. The face-detection model downloads on first run; the pack-wide startup dependency install (onnxruntime-gpu, mmdet, mmcv, modelscope, insightface…) applies regardless.