WanMoveTrackToVideo - ComfyUI Built-in Node Documentation

The WanMoveTrackToVideo node prepares conditioning and latent space data for video generation, incorporating optional motion tracking information. It encodes a starting image sequence into a latent representation and can blend in positional data from object tracks to guide the motion in the generated video. The node outputs modified positive and negative conditioning along with an empty latent tensor ready for a video model.

Inputs

Parameter	Description	Data Type	Required	Range
`positive`	The positive conditioning input to be modified.	CONDITIONING	Yes	-
`negative`	The negative conditioning input to be modified.	CONDITIONING	Yes	-
`vae`	The VAE model used to encode the starting image into the latent space.	VAE	Yes	-
`tracks`	Optional motion tracking data containing object paths.	TRACKS	No	-
`strength`	Strength of the track conditioning. (default: 1.0)	FLOAT	Yes	0.0 - 100.0
`width`	The width of the output video. Must be divisible by 16. (default: 832)	INT	Yes	16 - MAX_RESOLUTION
`height`	The height of the output video. Must be divisible by 16. (default: 480)	INT	Yes	16 - MAX_RESOLUTION
`length`	The number of frames in the video sequence, in increments of 4. (default: 81)	INT	Yes	1 - MAX_RESOLUTION
`batch_size`	The batch size for the latent output. (default: 1)	INT	Yes	1 - 4096
`start_image`	The starting image or image sequence to encode.	IMAGE	Yes	-
`clip_vision_output`	Optional CLIP vision model output to add to the conditioning.	CLIP_VISION_OUTPUT	No	-

Note: The strength parameter only has an effect when tracks are provided and strength is greater than 0.0; track conditioning is applied only when start_image is also provided. If tracks are not provided or strength is 0.0, the track blending is skipped. When track blending is active, the positive conditioning receives the track-blended latent image, while the negative conditioning receives the unmodified latent image. If start_image is not provided, no latent image and mask conditioning is created; the positive and negative conditioning pass through unchanged (except that clip_vision_output is still added if provided), and the node outputs an empty latent. Note: When start_image is provided, the image sequence is resized to the target width and height and truncated to the first length frames. If the sequence is shorter than length, the remaining frames are filled with neutral gray frames (value 0.5) before VAE encoding. The resulting conditioning includes a concat_mask with value 0 at the temporal positions corresponding to the start image frames and 1 elsewhere.

Outputs

Output Name	Description	Data Type
`positive`	The modified positive conditioning, potentially containing `concat_latent_image`, `concat_mask`, and `clip_vision_output`.	CONDITIONING
`negative`	The modified negative conditioning, potentially containing `concat_latent_image`, `concat_mask`, and `clip_vision_output`.	CONDITIONING
`latent`	An empty latent tensor with shape `[batch_size, 16, ((length - 1) // 4) + 1, height // 8, width // 8]`, determined by the `batch_size`, `length`, `height`, and `width` inputs.	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): b02a1a359d349a0136d84ed77a510c46cb2c8b565650ed54d5fca6c87cd0ab1f

​Inputs

​Outputs

Inputs

Outputs