LoadMediaPipeFaceLandmarker - ComfyUI Built-in Node Documentation - ComfyUI

Load Face Detection Model (MediaPipe)

Overview

This node loads a MediaPipe Face Landmarker v2 model, which can detect faces and facial landmarks (like eyes, nose, and mouth) in images. It contains two detection variants (short-range and full-range) along with shared mesh data, blendshapes, and canonical geometry for facial analysis.

Inputs

Parameter	Description	Data Type	Required	Range
`model_name`	Face detection model from models/detection/.	COMBO	Yes	List of available models in the `models/detection/` directory

Outputs

Output Name	Description	Data Type
`FACE_DETECTION_MODEL`	A loaded FaceLandmarker model object containing both detection variants (short/full), connection sets for facial topology, canonical data, and model patchers for GPU management.	FACE_DETECTION_MODEL

Note: The output is a complex object that can be used by other nodes for face detection and landmark extraction tasks. It contains two detection variants: “short” for close-range detection and “full” for full-range detection.

This documentation was AI-generated. If you find any errors or have suggestions for improvement, please feel free to contribute! Edit on GitHub

Source fingerprint (SHA-256): 33dda845b572ccffc1bd4b64fb9c338ce4313783b092fe311d89741a211f18c9