Convert to Batch

ComfyUI has a quiet type system fight going on, and you will run into it the first time you process a folder of images: some nodes hand you a list of images, and the next node in line insists on a batch - a single tensor with a dimension for "how many". A list is Python's thing; a batch is torch's thing. They look the same on screen and are absolutely not the same under the hood. Convert to Batch is the bridge.

What it actually does

The Eclipse To Batch [Eclipse] node takes a list of values and combines them into a batch format. It's smart about what kind of list you fed it, because the schema uses a dynamic match type (COMFY_MATCHTYPE_V3) instead of forcing you to pick "Image", "Mask", "Latent" etc. from a dropdown first. Feed it a list of images and you get an image batch tensor. Feed it masks, latents, or conditioning and it combines those. Feed it plain values - strings, numbers - and it merges them into a single comma-separated string, which is a handy trick on its own: turn five per-image labels into one text block without writing a join node.

That no-dropdown behavior is the part worth calling out. A lot of pack-specific "to batch" nodes make you pick the type manually and throw an error if you guess wrong. This one looks at the wire and figures it out, so it slots into a workflow without any configuration.

Why you'll reach for it

The classic trigger: you load several images (Eclipse's Load Image From Folder or Load Batch From Folder output lists), process each one through per-item nodes that emit lists, and then hit a node like a save-composite or an upscaler that needs a single IMAGE batch. Rather than re-architecting the whole graph, you bolt a Convert to Batch on the end. It's also the natural pair to Convert to List - one splits a batch apart, the other stitches a list back together, and together they let you "open up" a batch to process items individually and then rejoin cleanly.

Inputs and outputs

One input, input: any list of values or tensors you want combined. One output, the combined batch. There are no widgets to fiddle with beyond that - if you're getting an error, it's almost always because the upstream list is empty (the node can't batch zero items) or you connected a single item where a list was expected. Check what the node before you is actually emitting; "it's a list, trust me" is the usual culprit.

Install it

It ships in ComfyUI_Eclipse, so install once for the whole family. ComfyUI Manager → search "Eclipse" → install, then restart. Or:

cd ComfyUI/custom_nodes
git clone https://github.com/r-vage/ComfyUI_Eclipse

The pack's requirements (opencv-python, pilgram, PyYAML, aiohttp, plus the standard torch/numpy/Pillow) are covered for most installs; portable setups may need to run pip install -r custom_nodes/ComfyUI_Eclipse/requirements.txt with the embedded Python. You'll find the node under Eclipse → Conversion.

Gotchas

The comma-separated primitives behavior surprises people - if you batch strings and expected a list of strings out, this node gives you one combined string instead. That's by design, and it's genuinely useful for labels or filenames, but know it going in. The other thing to remember: a list and a batch behave differently downstream, so converting a batch to a list and back doesn't always reproduce the exact same shape. Use it for what it's for - getting a list into the format the next node actually wants - and you're golden.