Batch Sequence (mtb)

If you've generated a video in chunks - a batch for the intro, a batch for the middle section, a batch for the outro - Batch Sequence is how you glue them back into one continuous batch, in order, as a straight hard cut with no blending. Think of it as playlist logic applied to tensors: instead of one long batch, you feed it several shorter ones and it plays them out one after another.

Where it fits

This is part of MTB Nodes' mtb/batch family, and it's the plainer sibling of Batch Sequence Plus - that one adds crossfades and wipes between segments, this one just concatenates. Reach for Batch Sequence when a hard cut between segments is actually what you want (different scenes, a deliberate jump cut), and reach for the Plus version when you want the join to feel smooth.

The input that matters - and the one that isn't shown

The declared schema only has one field: reverse (default false), which plays the assembled sequence backwards once it's built. There's no explicit list of image-batch inputs in the schema, and that's not an oversight - it's the same "dynamic inputs" pattern the pack's own Debug node calls out by name: you connect batches by wiring them in, and ComfyUI grows new input slots on the node as you go, rather than the node declaring a fixed number of sockets up front. So in the graph, you'll see extra unlabeled inputs appear as you connect more segments - that's expected, not a bug.

Output is a single IMAGE - all the input batches, concatenated along the frame dimension in the order you wired them (or reversed, if you flipped that toggle).

Installing it

ComfyUI Manager: search MTB Nodes, install, restart. Or:

cd ComfyUI/custom_nodes
git clone https://github.com/melMass/comfy_mtb

then restart ComfyUI. No models, no heavy dependencies - this is tensor concatenation.

Common issues

The thing that actually breaks this node in practice is resolution mismatch: if your segments aren't the same width and height, concatenation along the batch dimension fails outright, because unlike its sibling Concat Images (which has an on_mismatch option to resize to the smallest or largest input), Batch Sequence has no such fallback in its schema - so make sure every batch you feed it is already the same size before it hits this node, or resize upstream with a standard image-resize node.

The other issue is pack-wide: MTB logs [comfy_mtb] Some nodes (N) could not be loaded at startup instead of crashing outright when one of its many nodes can't import, and points you at http://127.0.0.1:8188/mtb for details - a real, confirmed behavior from other users' install logs. If Batch Sequence isn't showing up in your node search after installing the pack, that startup line is where to look before assuming the clone failed.