NormalizeVideoLatentStart - ComfyUI Built-in Node Documentation

This node adjusts the first few frames of a video latent to make them look more like the frames that come after. It calculates the average and variation from a set of reference frames later in the video and applies those same characteristics to the starting frames. This helps reduce differences between the starting frames and the rest of the video, creating a smoother and more consistent transition.

Inputs

Parameter	Description	Data Type	Required	Range
`latent`	The video latent representation to process.	LATENT	Yes	-
`start_frame_count`	Number of latent frames to normalize, counted from the start (default: 4).	INT	Yes	1 to 16384 (max resolution)
`reference_frame_count`	Number of latent frames after the start frames to use as reference (default: 5).	INT	Yes	1 to 16384 (max resolution)

Note: The reference frames are taken starting right after the start_frame_count frames. If there are fewer frames available than reference_frame_count requests, the node uses however many are available (at most one less than the total frame count of the latent). If the video latent has only 1 frame, no normalization is performed and the original latent is returned unchanged.

Outputs

Output Name	Description	Data Type
`latent`	The processed video latent with the starting frames normalized.	LATENT

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): 383e5a19ee4cd8bdea5983567ddbdc30bb09c373142a1a934cea985f1b9d1b0d

​Inputs

​Outputs

Inputs

Outputs