Loop Strip — Face-Aligned Center

A ComfyUI node in LoopStrip with 8 inputs and 3 outputs.

By serhiiyashyn-sf·Created 4 months ago·Updated 4 months ago· 0

◄canvas_size1024►

◄face_fill0.18►

◄face_y0.35►

◄use_face_detectionfalse►

◄yolo_modelface_yolov8m.pt►

◄confidence0.35►

CategoryLoopStrip

Inputs (8)

Name	Type	Default	Description
images	IMAGE		—
canvas_size	INT	102464–8192	Square output size (pixels).
face_fill	FLOAT	0.180.02–0.9	Zoom knob. Face height as a fraction of the canvas. LOWER = zoom out (smaller character, more empty space). HIGHER = zoom in. Same value across every call = same face size for every character sheet.
face_y	FLOAT	0.350–1	Vertical position of the face center on the canvas (0=top, 1=bottom). 0.35 puts the face in the upper-third.
use_face_detection	BOOLEAN	false	OFF (default): always use the silhouette algorithm — head-cx detection from the top of the trimmed body, batch-consistent size. Produces clean results on chibi characters with white BG. ON: also try YOLO face detection where available; batch-median face height drives scale, silhouette head-cx is used as fallback for back views. Only turn on if you've verified face YOLO works well on your character style.
yolo_model	COMBO	face_yolov8m.pt	Face-detection model (only used when use_face_detection is ON). Looked up in ComfyUI/models/ultralytics/{,bbox}/...
confidence	FLOAT	0.350.05–0.95	YOLO confidence threshold (only used when use_face_detection is ON). Lower if sides/3-4 angles aren't being detected.
maskopt	MASK		Optional foreground mask; overrides white-threshold detection.