Audio Sequence (mtb)

As ComfyUI has grown into audio and video work, you sometimes just need to glue clips together on a timeline - a bit of narration, then a sound effect, then more narration. Audio Sequence does that: it concatenates multiple audio inputs one after another into a single clip, with control over the gap (or overlap) between them.

It's one of MTB's newer audio utilities, and it's the "put these in a row" node. Its sibling, Audio Stack, layers clips on top of each other; this one lines them up end to end.

How it works

You wire in several audio clips (the node takes dynamic inputs, so you add as many as you need), and it stitches them in order. Between each clip it inserts silence_duration seconds of silence. Make that value negative and the clips overlap instead - and the node safely bounds the overlap to the clip length so you can't crossfade past the edges. It also does the housekeeping automatically: everything gets resampled to the highest sample rate among the inputs, and if any input is stereo, they all get converted to stereo so the output is consistent.

The inputs and outputs that matter

• silence_duration (FLOAT, default 0, range −999 to 999) - the one knob. Positive adds a gap between clips; 0 butts them together; negative overlaps them.

You connect the clips themselves as inputs. The output is a single sequenced_audio (AUDIO) - the assembled timeline.

How to install it

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

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

then restart. This is an audio node, so it leans on ComfyUI's audio stack (torchaudio); if audio nodes across your install are misbehaving, that dependency is usually the thing to check.

Common issues & troubleshooting

The clips play on top of each other, not in sequence. You've probably grabbed Audio Stack by mistake - that's the overlay node. Sequence is the end-to-end one. (Or your silence_duration is a large negative number, overlapping everything.)

The overlap isn't crossfading the way I hoped. Negative silence_duration overlaps the raw audio; it's bounded to the clip length but it's a plain overlap, not a shaped crossfade. For a real fade you'd shape the clips first.

Sample rate / channel mismatches. Not something you have to fix by hand here - the node resamples to the highest input rate and promotes everything to stereo if any clip is stereo. Just know that a mono clip alongside a stereo one comes out stereo.