Face Tools Nodes

The Face Tools nodes detect faces with MediaPipe and provide utilities for identifying, masking, and extracting faces in workflows. The current nodes are in the segmentation category and use version 1.2.2.

FaceIdentifier

FaceIdentifier outputs a copy of the input image with detected face IDs printed on each face. Use it first when you need to target a specific face with FaceMask or FaceOff.

Face IDs are numbered from 0. Detection order can change if the image changes, so run FaceIdentifier again after editing an image.

Inputs

Input	Description
Image	Image to face detect.
Minimum Confidence	Minimum confidence for face detection. Lower this if detection is failing.
Chunk	Bypass full-image face detection and use chunking. Chunking is also used automatically if no faces are found in the full image.

Outputs

Output	Description
Image	The input image with face ID numbers drawn on detected faces.
Width	Output image width in pixels.
Height	Output image height in pixels.

FaceMask

FaceMask creates a mask for detected faces on the input image.

Leave Face IDs empty to mask all detected faces, or provide a comma-separated list such as 0,2,7 to target specific faces. Use FaceIdentifier to find the IDs.

The mask can be adjusted with X and Y offsets if detection is slightly too large or small. Enable Invert Mask to affect everything except the detected faces.

Inputs

Input	Description
Image	Image to face detect.
Face IDs	Comma-separated list of face IDs to mask. Leave empty to mask all detected faces.
Minimum Confidence	Minimum confidence for face detection. Lower this if detection is failing.
X Offset	Offset for the X-axis of the face mask.
Y Offset	Offset for the Y-axis of the face mask.
Chunk	Bypass full-image face detection and use chunking. Chunking is also used automatically if no faces are found in the full image.
Invert Mask	Toggle to invert the mask.

Outputs

Output	Description
Image	The original image, converted to RGBA.
Width	Output image width in pixels.
Height	Output image height in pixels.
Mask	The generated face mask.

FaceOff

FaceOff extracts a single detected face into a bounded image and returns a matching mask plus paste coordinates. Use FaceIdentifier to find the face ID before targeting a specific face.

Padding expands the bounding box around the detected face. This gives downstream processing more context and increases the bounded image size while keeping the face in place within the crop.

Inputs

Input	Description
Image	Image for face detection.
Face ID	The face ID to process, numbered from 0. Multiple faces are not supported.
Minimum Confidence	Minimum confidence for face detection. Lower this if detection is failing.
X Offset	X-axis offset of the mask.
Y Offset	Y-axis offset of the mask.
Padding	All-axis padding around the mask in pixels.
Chunk	Bypass full-image face detection and use chunking. Chunking is also used automatically if no faces are found in the full image.

Outputs

Output	Description
Image	The bounded face image. If no face is found, the original image passes through.
Width	Output image width in pixels.
Height	Output image height in pixels.
Mask	Mask matching the bounded image.
X	X coordinate of the bounding box’s left side.
Y	Y coordinate of the bounding box’s top side.

Tips

• Use the same Minimum Confidence value in FaceIdentifier and the FaceMask or FaceOff node that consumes the IDs.
• Enable Chunk if not all target faces are detected. Full-image detection and chunked detection can produce different results.
• Lower Minimum Confidence when detection fails, but watch for false positives.
• Adjust X and Y offsets if the mask is too large, too small, or shifted.
• Add FaceOff padding when the extracted face needs more surrounding context.
• Face detection can fail on heavy face paint, hair covering the face, extreme angles, or other obstructions.