⚠ Convert to List (Deprecated)
The deprecated batch-splitter, and its smarter replacement
- input
- output
If you're on this page because a saved workflow has a red node with "Convert to List" on it, here's the short version: it's deprecated, Eclipse v4.0.0 deleted the legacy nodes, and there's a migration tool that fixes your workflow in one command. The longer version - what the node did and what replaced it - is below.
What it was
The inverse of Convert to Batch. Where the batch converter stacked a list of images into one tensor, this one split a batched tensor back into a list of individual items. The default mode was IMAGE_BATCH_TO_LIST: feed it a [N, H, W, C] image batch and it returned a Python list of single-image tensors, so downstream nodes that iterate per-image could chew through the frames one at a time. It handled masks and nested lists the same way, and it had a fallback - if you fed it something that wasn't a batch, it wrapped it as a one-item list instead of crashing.
That per-item split matters in real workflows: image-comparer loops, per-frame video processing, anything that wants "give me one image at a time, please." But the dropdown-based design (convert_to) was clunky, and the pack moved on.
What replaced it
The current node is Convert to List (To List [Eclipse]), still in Eclipse → Conversion. It auto-detects the input type - Image, Mask, Latent, Conditioning, whatever - and outputs a list of individual items with no dropdown to get wrong. Same job, less ceremony. The description says it plainly: "Requires no dropdown selection."
For old workflows, don't rebuild by hand:
python tools/migrate_workflow.py <path_to_workflow.json_or_directory>
The script creates backups before rewriting node IDs, so it's safe to point at a whole folder of workflows.
How the legacy node worked
The batch-splitting logic walked the input, recursing through nested lists/tuples, and for each 4D tensor it sliced off every batch item (item[i:i+1]) into its own list element. 3D tensors got a leading dimension added, and numpy arrays went through the same path. That's a faithful split, not a copy - each list entry is a slice of the original.
A note on the pack itself
This is part of ComfyUI_Eclipse, the rewritten successor to RvTools. If you're migrating a workflow from the RvTools era, the migration tool is your friend - the old repo even disappeared from GitHub briefly in early 2025, so a lot of old workflows are running on node IDs that no longer exist anywhere except the migration map.
Install
cd ComfyUI/custom_nodes
git clone https://github.com/r-vage/ComfyUI_Eclipse
Restart ComfyUI. Then replace the legacy node with the auto-detect To List variant (or run the migration script).
Troubleshooting
- Red node / "missing node type": expected post-v4.0.0. Migrate.
- List comes back with one big tensor instead of separate items: you're probably on the legacy node with the wrong
convert_tomode, or feeding it a batch it treats as already-split. The auto-detect current node avoids this. - Per-frame processing skips frames: check whether you're splitting a list of lists - the recursion nests correctly, but a workflow that expects flat lists can be surprised by one level of nesting.
Inputs (2)
| Name | Type | Default | Description |
|---|---|---|---|
| input | * | Any value to convert to list format | |
| convert_to | COMBO | IMAGE_BATCH_TO_LIST | Target list conversion type |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| output | * | — |