Audio List to Batch

The name is the whole job description. The pack's Audio Segmenter hands you a list of clips; most of the rest of ComfyUI's audio world wants a single batched AUDIO tensor. This node is the bridge between those two worlds, and it's the node you reach for the moment you want to feed split clips into anything that expects batch input, like an ASR or TTS node.

How it works

Mechanically it's simple and honest: it takes the audio_list, pads every clip with silence (zeros) so they all match the length of the longest clip, then stacks them along the batch dimension into one AUDIO object. Same sample rate as the source, same shape conventions, just batched. There's no resampling and no clever stitching - if the clips came out of one Audio Segmenter pass, they already share a sample rate, so the whole thing is lossless apart from the zero-padding.

Inputs and outputs

It has exactly one input and one output, which is why this is a 30-second learn:

• audio_list - the list of clips from Audio Segmenter (or anything else producing that shape). It's typed as *, a loose wildcard, so ComfyUI won't shout at you if you wire in the wrong thing - it'll just produce garbage or an empty fallback at runtime.
• AUDIO - one batched audio object, padded to the longest clip.

One honest warning about the padding: it's silent filler, not trimming. If your clips are wildly different lengths, your batch is only as "dense" as the longest clip, and downstream nodes will chew through that silence. For most ASR batch passes that's fine; if you're memory-sensitive, split with a tighter max_length_s upstream instead.

Installing the pack

ComfyUI Manager: search "Audio Segment" (pack title ComfyUI Audio Segment Plugin). Manual:

cd path/to/ComfyUI/custom_nodes/
git clone https://github.com/ahkimkoo/Comfyui-AudioSegment
cd Comfyui-AudioSegment
pip install -r requirements.txt

Then restart ComfyUI - nodes land under Audio/Segmentation. Note the pack's requirements are heavy (torch, openai-whisper, transformers, accelerate, pydub, opencc-python-reimplemented) because the pack shares one install across five nodes; ComfyUI already ships torch, so watch that a fresh pip run doesn't churn your versions.

Gotchas

Feed it an empty list and it returns a silent one-sample fallback rather than erroring - useful for lazy graphs, slightly confusing if you were expecting a loud failure. And remember it pads, it doesn't trim: your batched clips all run to the length of the longest. If you're here because the paragraph node's AudioClip list doesn't plug in, that's expected - this node pairs with the plain Audio Segmenter list, while the paragraph node's clips go into Select AudioClip From List instead. Pick the pairing that matches your splitter.