LatentCutToBatch - ComfyUI Built-in Node Documentation

The LatentCutToBatch node splits a latent representation along a chosen dimension (time, width, or height) into slices of a specified size and stacks them into a new batch. Each slice becomes a separate item in the batch, so different parts of a latent sample can be processed independently.

Inputs

Parameter	Description	Data Type	Required	Range
`samples`	The latent representation to be split and batched.	LATENT	Yes	-
`dim`	The dimension along which to cut the latent samples. `"t"` refers to the temporal (frame) dimension, `"x"` to the width, and `"y"` to the height.	COMBO	Yes	`"t"` `"x"` `"y"`
`slice_size`	The size of each slice to cut from the specified dimension. If the dimension’s size is not perfectly divisible by this value, the remainder is discarded. (default: 1)	INT	Yes	1 to 16384 (max resolution)

Note: The "t" option only takes effect when the latent includes a temporal dimension. If the chosen dimension maps to the batch or channel position, or does not exist (for example, selecting "t" on a latent without frames), the node returns the input unchanged. If slice_size is larger than the size of the chosen dimension, the entire dimension is used as a single slice. When the dimension size is not evenly divisible by slice_size, the leftover portion at the end is discarded. The output batch size is the input batch size multiplied by the number of slices, and the sliced dimension itself is reduced to slice_size.

Outputs

Output Name	Description	Data Type
`samples`	The resulting latent batch, containing the sliced and stacked samples.	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): 873c9bc8391971887f1ab636c086cab86f5504a9c653bc80b54120ee53980bdf

​Inputs

​Outputs

Inputs

Outputs