Insert Latent To Index(InsertLatentToIndex)

The Insert Latent To Index node receives two inputs of type LATENT and supports both 4D (B, C, H, W) and 5D (B, C, T, H, W) tensor formats.

The Insert Latent To Index node is used to insert a batch of latent representations into a specified position within another latent batch. Its core function is to replace the latent sample at a specific index in the target batch, enabling precise editing and composition within the latent space.

Node Functionality

This node receives two inputs of type LATENT and supports both 4D (B, C, H, W) and 5D (B, C, T, H, W) tensor formats. Internally, it uses PyTorch's torch.cat function to concatenate the source tensor before the specified index of the destination tensor, skipping the original data at that index in the destination tensor, and finally outputs a new combined latent tensor.

Node Parameter Description - Insert Latent To Index

Connection Inputs (Inputs)

Control Parameters (Parameters)

Output

Usage Scenarios

In an image generation workflow, this node can be used to replace the latent representation of a specific image within a batch generation. For example, after generating a set of portrait images, if you want to replace the facial features of the third image with a more satisfactory one, you can use the new facial latent representation as source, the original batch as destination, and set index to 2 to complete the replacement.

Notes

When using this node, note that the specified index value must be within the valid range of the target latent batch; otherwise, the node will report an error. For 4D tensors, the index corresponds to the batch dimension (B); for 5D tensors, the index corresponds to the time dimension (T).

Insert Latent To Index Node Source Code Link

The Insert Latent To Index node is from the ComfyUI-KJNodes node package.