Merge audio list to master audio

Chunked generation produces chunked audio. You render your video in segments to fit VRAM, each segment gets its own little audio clip, and now you've got a pile of clips that need to be one continuous soundtrack before you save anything. Merge audio list to master audio is the node that concatenates them - with the details handled so the joins don't pop.

The key thing to understand before you reach for it: the input is a list, not a single clip. The node sets INPUT_IS_LIST = True internally, which means it collects every audio wire you feed into that socket on a given execution and merges them all in one pass, instead of looping one clip at a time. The tooltip says it plainly: "a list of audio to convert to 1 audio stream." If you've been wiring audio one clip at a time and wondering why nothing merges, that's the disconnect.

How it works

It takes the list of clips, grabs the sample rate from the first one, and concatenates the waveforms end to end. The part that shows care: before each join it applies a tiny 0.1-second fade - fading the tail of the previous clip out and the head of the next clip in - so you don't get the click or pop that a hard concatenation produces at every boundary. There's a safety check that shrinks the fade if a clip is too short to hold it, so tiny clips don't error out.

It's written to sit after exactly the kind of workflow the README describes: concatenating a bunch of separately-generated videos (the comment name-checks VHS-style combine) where each one dragged its own audio along. Merge the audios with this, merge the images with a video combiner, and you've got a master pair.

The input that matters

Just one: audio_list (AUDIO). That's the whole interface. Feed it the list of clips you want in order - order matters, because this is concatenation, not mixing. (If you want to overlay clips on top of each other instead of string them together, that's TKAudioFuse, a different job entirely.)

Output

One AUDIO socket: the master track, flattened to a single batch. Wire it into your final save/combine step.

Common issues

The big one is feeding clips that don't share a sample rate - the node takes the rate from the first clip and concatenates the rest as-is, so mismatched rates give you audio that speeds up or slows down mid-track. Resample everything to one rate before merging. Second gotcha: it joins whatever order you give it, so keep your list sorted by timeline position or you'll get the movie with scenes in the wrong order. And if you expected crossfades long enough to be musical, the 0.1s fade is purely a pop-killer - it's not a transition tool.

Installing it

Part of trashkollector/TKNodes ("ComfyUI Handy Nodes"). Install via ComfyUI Manager (search "Handy Nodes") or:

cd ComfyUI/custom_nodes
git clone https://github.com/trashkollector/TKNodes

Restart, and it's under the HandyNodes-KT category. No models; the pack's standard torchaudio + pydub requirements apply.