EmptyHunyuanVideo15Latent - ComfyUI Built-in Node Documentation - ComfyUI

This node creates an empty latent tensor specifically formatted for use with the HunyuanVideo 1.5 model. It generates a blank starting point for video generation by allocating a tensor of zeros with the correct channel count and spatial dimensions for the model’s latent space.

Inputs

Parameter	Description	Data Type	Required	Range
`width`	The width of the video frame in pixels.	INT	Yes	-
`height`	The height of the video frame in pixels.	INT	Yes	-
`length`	The number of frames in the video sequence.	INT	Yes	-
`batch_size`	The number of video samples to generate in a batch (default: 1).	INT	No	-

Note: The spatial dimensions of the generated latent tensor are calculated by dividing the input width and height by 16. The temporal dimension (frames) is calculated as ((length - 1) // 4) + 1. These calculations use integer division, so width and height should be multiples of 16 to avoid truncation.

Outputs

Output Name	Description	Data Type
`samples`	An empty latent tensor with dimensions suitable for the HunyuanVideo 1.5 model. The tensor has a shape of `[batch_size, 32, frames, height//16, width//16]`. The output also includes a `downscale_ratio_spacial` value of 16.	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): ce7ec75e8433c778d175a3e2ea260a4397aa5507428908b9a32f50fbe9e184c6