🌀 BV ImagePipe Loader

If you've ever hand-fed fifty images into ComfyUI one at a time to build a training set, you know the pain this node removes. BV ImagePipe Loader is the front door of the BV Nodes batch pipeline: point it at a directory and it turns everything in there into a single BV_IMAGE_PIPE, complete with a count for every image so the rest of the graph can track it.

The pipe concept is the same trick rgthree's Context nodes popularized - bundle a bunch of data down one wire instead of dragging five separate connections across the canvas. Here the bundle is (image, caption, count), and the Loader is where those counts get assigned. The splitter, junction, merger, and caption saver in this pack all speak the same pipe, so the Loader is usually the first node you place.

What you actually set

• directory (required) - the folder to scan. Absolute path; ComfyUI won't guess relative to your workflow.
• zero_padding (default 3) - leading zeros in the count, so 3 gives you 001, 010, 100. Turn it to 0 for bare numbers.
• image_load_cap (default 0) - stop after N images; 0 means no limit.
• start_index (default 0) - where numbering begins. It's the offset that feeds every downstream %count placeholder, so bump this if you're processing a set in chunks and want the numbers to continue, not restart.

What comes out

• IMAGE_PIPE (list) - everything the rest of the BV nodes consumes.
• FILE PATH (list) - the full path of each loaded image, one per item. Handy for wiring into a captioner that wants a filename, or for logging.
• PATH (string) - the directory you gave it, echoed back for convenience.

The image order is whatever the folder scan returns, which is why the counts matter more than the sort: the merger at the other end reconstructs order by count, not by directory listing.

Install

ComfyUI Manager is the easy route - search "BV Nodes" (the pack title) and install. Or clone it by hand:

cd ComfyUI/custom_nodes
git clone https://github.com/BlackVortexAI/ComfyUI-BVortexNodes.git

Restart ComfyUI. There are no pip dependencies and no model files - the whole pack is plain ComfyUI nodes, so this is a two-minute install, not a download-a-checkpoint afternoon.

Where people trip

The two numbers that bite are start_index and image_load_cap - if the set looks short or the counts start somewhere you didn't expect, do the arithmetic on those before suspecting a bug. And point the loader at the folder itself, not a file; a path to one image will just load that one image, which is technically correct and deeply unhelpful.