SAM3_VideoTrack - ComfyUI Built-in Node Documentation - ComfyUI

Track objects across video frames using SAM3’s memory-based tracker. The node processes a sequence of video frames and maintains object identities across frames, using either initial masks or text prompts to define what to track, and can detect new objects along the way using text conditioning.

Inputs

Parameter	Description	Data Type	Required	Range
`images`	Video frames as batched images	IMAGE	Yes	Batched video frames
`model`	The SAM3 model to use for tracking	MODEL	Yes	SAM3 model
`initial_mask`	Mask(s) for the first frame to track (one per object)	MASK	No	One mask per object
`conditioning`	Text conditioning for detecting new objects during tracking	CONDITIONING	No	Text conditioning
`detection_threshold`	Score threshold for text-prompted detection (default: 0.5)	FLOAT	No	0.0 to 1.0 (step 0.01)
`max_objects`	Max tracked objects. Initial masks count toward this limit. 0 uses the internal cap of 64. (default: 4)	INT	No	0 to 64
`detect_interval`	Run detection every N frames (1=every frame). Higher values save compute. (default: 1)	INT	No	1 or higher

Note: Either initial_mask or conditioning must be provided. If both are omitted, the node raises an error.

Outputs

Output Name	Description	Data Type
`track_data`	Tracking data containing object masks and metadata across all video frames	SAM3_TRACK_DATA

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): ef584628b334997a001a857a7deffb7eda34db8fa50e3d734a07b5e92566d48d