Image Batch Assembler

This is the node the whole pack quietly depends on. Image Batch Assembler takes any number of image inputs - from LoadImage nodes, from generation, from anywhere - and hands you one clean output to feed into Multi Image Inserter. Without it you'd be manually rewiring half a dozen images into every insertion, which is exactly the pain this pack exists to remove.

It's from ComfyUI Image to Video Inserter by Yeq6X, and it's one of the three nodes the README actually documents. The workflow it anchors is: collect your stills here, decide how they should be shaped, then insert them at specific frame positions in a video. The repo's example workflow does precisely that - blank frames in, a few loaded images assembled, straight into the inserter.

How it works - the part that surprises people

The clever bit is the dynamic inputs. Under the hood the node is declared with inputs up to image_1000, but you control how many actually show with inputcount, then click the "Update inputs" button on the node to add or remove image sockets. Set inputcount to 4, hit the button, and you get image_1 through image_4. The button lives in the pack's custom frontend (the web/ folder it ships), so it only appears if the pack loaded properly.

The second decision is output_type:

• list - returns the images as a Python list, keeping each one at its own size. This is the mode Multi Image Inserter expects.
• tensor - concatenates everything into one B,H,W,C batch. That requires a uniform size, which is what resize_mode is for.

Inputs that matter

• inputcount - how many image inputs exist (1–1000, default 2). Change it, click "Update inputs".
• output_type - list or tensor. List for this pack's inserter; tensor for anything expecting a plain batch.
• resize_mode - only used in tensor mode, and only when sizes differ:
  • pad - no resizing, center-pad each image to the largest size with black borders.
  • fit_short_side - shrink to fit the target box, letterboxed (black bars).
  • cover_fill - scale up to cover the target and center-crop the excess. This is the "no bars" option and usually looks best for video.

The output is images (IMAGE). In list mode it's a list of single-frame tensors; in tensor mode a real batch.

Installing it

Standard pack install:

cd ComfyUI/custom_nodes
git clone https://github.com/Yeq6X/ComfyUI-image-to-video-inserter

then restart ComfyUI, or use ComfyUI Manager and search "Image to Video Inserter". No models, no keys. If the node appears but the "Update inputs" button doesn't, your ComfyUI is loading a stale frontend - a hard refresh usually sorts it.

When to reach for it

Reach for it any time you're feeding multiple stills into one operation - assembling a batch of images to insert into a video, or just gathering several loaded images into a single wire so you can stop stacking repeat nodes. One honest caveat: it returns a list, not a standard batch, in its default mode, so if you plug it into a node that expects a plain IMAGE tensor you'll want output_type set to tensor instead. For this pack's own inserter, though, it's the exact right shape - and it's the reason Multi Image Inserter stays so simple to use.