Vantage Unbatch Images

Vantage Unbatch Images is the reverse of all the batch-building nodes: it takes one IMAGE batch and splits it back into individual images, up to eight separate outputs. One tensor in, up to eight single-image wires out.

You need this more often than you'd think. A lot of ComfyUI workflows produce a batch - AnimateDiff frames, a batch run at a given size - and then every node downstream wants to touch the images individually. Maybe you want to face-detail each frame, save each one with its own filename, or run different post-processing on each. The built-in graph model doesn't make that trivial; this node does it mechanically.

How it works

run() simply indexes the batch dimension:

• image_1 = the first image in the batch (images[0])
• image_2 = the second, and so on up to image_8

Each output is a single-image batch (shape [1, H, W, C]), so it wires straight into any node that expects an IMAGE - a preview, a save, an upscaler, a face-detailer. Slots beyond the actual batch size come out as None: feed it a 4-frame batch and image_5 through image_8 are all None.

Input:

• images (IMAGE) - "Image batch to unbatch."

Outputs: image_1 through image_8, each an IMAGE.

Where you'd reach for it

• AnimateDiff / video frames: decode a batch of frames, split them, and save or post-process each frame on its own path.
• Batch runs: generate 4 images in one queue, then unbatch so each can take a different upscaler or LoRA-based refinement.
• Grid-free saving: when you want individual files instead of one combined image.

It pairs cleanly with the pack's Append Image Batch on the other end - split, process each piece, append them back together.

Gotchas

• None outputs. Slots beyond the real batch size are None, and most nodes error on None. Wire only the outputs you know will exist, or gate the extras behind a Conditional Pass Through.
• Eight is the ceiling. A 12-frame batch will happily give you the first 8 and silently swallow the rest. If you regularly work with larger batches, you'll want a loop-style splitter or a list-based approach instead of this node.
• Each output is still a batch of one, so a downstream node that flattens a single-image batch will work fine - but the shape is [1, H, W, C], not [H, W, C].

Installation

Part of Vantage Nodes - install the pack once, use every node. ComfyUI Manager: search "Vantage Nodes." Or:

cd ComfyUI/custom_nodes
git clone https://github.com/vantagewithai/Vantage-Nodes.git
pip install -r requirements.txt

Restart ComfyUI. No model files involved for this node.

Troubleshooting

• Some outputs are empty/None: that's the batch being shorter than 8, working as designed. Count your frames.
• First image is right but the rest are wrong order: the node preserves batch order - image_1 is truly the first frame. If your batch came from a sampler that reorders, fix it upstream, not here.