Pipe Batch Index

Pipe Batch Index does one thing: it slices a chunk out of a batch that's riding inside an Easy-Use pipe. If you ran a batch of 16 latents and only want images 4 through 8 to go on to a detailer or an upscale pass, this is how you cut that slice out without splitting your whole graph into two paths.

Why a "pipe" in the first place. ComfyUI-Easy-Use descends from TinyTerraNodes' idea of bundling model, conditioning, latent, VAE, CLIP and image into one PIPE_LINE wire instead of five or six separate spaghetti connections. It's the same instinct behind the "Efficiency Nodes" pack and part of why people reach for Easy-Use in the first place - fewer wires, easier to read at a glance. easy pipeBatchIndex is a small utility that operates on that bundle rather than on a raw LATENT socket, which is the whole point of staying in pipe-land: you don't have to break the pipe apart, slice the latent, and reassemble everything just to work on part of a batch.

Inputs. Three, all required: pipe (the bundle you're slicing from), batch_index (0 to 63, default 0 - where the slice starts), and length (1 to 64, default 1 - how many items you're pulling). So batch_index=4, length=4 grabs items 4-7 out of your batch. There's no optional input here - this node assumes you already have a populated pipe with a latent batch sitting inside it.

Output. Just pipe - the same bundle, but with its latent now trimmed to the slice you asked for. Wire that straight into whatever comes next: an easy pipeOut if you want the pieces broken out individually, a detailer node, a save node, or another presampling stage.

Install. Through ComfyUI Manager, search "ComfyUI Easy Use," or manually:

cd ComfyUI/custom_nodes
git clone https://github.com/yolain/ComfyUI-Easy-Use

then restart. This node has zero extra dependencies of its own - it's arithmetic on tensor indices, nothing more - so if the base pack installed cleanly, this node works out of the box.

Common issues. The one real trap is going out of bounds: if your actual batch size is smaller than batch_index + length, you're asking for latents that don't exist. Keep the math honest against whatever batch size you set upstream (in your loader or your presampling node), and remember batch_index is zero-based - item "4" is the fifth image in the batch, not the fourth. Beyond that, this is about as low-risk as an Easy-Use node gets: it's a slicing utility, not something that touches a model or pulls in a dependency that can break on you.