Loom Audio In

Audio is the odd one out in ComfyUI - most workflows never touch it, and the ones that do (video generation, voice, music) suddenly have a wire type they've never organized before. Loom Audio In bundles an AUDIO value into a LOOM stream, which is more useful than it sounds once your graph has multiple audio stages.

Why you'd reach for it

The c4f-wire-loom README explicitly lists AUDIO among the supported types, and it's there for a reason: video pipelines often generate audio in one stage, process it in another, and merge it back with the final render. Bundle the audio into the loom at the front, pull it out where it needs processing, push the result back in, and the final export stage picks up the current version. Same stream semantics as everything else in the pack - downstream sees the latest, not a stale copy.

It also keeps audio from being the one wire you have to route around every video node. In a graph that's otherwise all IMAGE and LATENT, the lone AUDIO connection is exactly the spaghetti strand looms were invented to kill.

How it works

The standard typed loom insert:

key = f"AUDIO_{label}" if label else "AUDIO"
new_loom[key] = audio
return (new_loom,)

Empty label → AUDIO; a label like voiceover → AUDIO_voiceover. The matching Loom Audio Out uses the same label. Drag an audio slot onto a collapsed Loom Split hub and the frontend auto-creates the node with the label filled in.

The inputs that matter

• audio (AUDIO): from an audio loader, a video model's audio output, or audio-processing nodes.
• label (string): the key. Name distinct audio streams (background vs. dialogue) so they don't collide.
• loom (LOOM, optional): the bundle to extend.

Output is a single LOOM, wired into Loom Split or Loom Join, with the audio pulled back out via Loom Audio Out at each processing or export stage.

Installing

Part of c4f-wire-loom. ComfyUI Manager → search "c4f-wire-loom" → install → restart, or:

cd ComfyUI/custom_nodes
git clone https://github.com/Codes4Fun/c4f-wire-loom

No extra dependencies - importantly, the pack ships no audio libraries of its own; whatever handles AUDIO is already in your ComfyUI, and this node just routes the objects.

Where it bites

The usual loom rules hold: label your audio streams or an unlabeled second one overwrites the first, and only subgraphs downstream of the Loom In see it. One audio-specific note - if your audio pipeline is split between a video model's output and a separate audio tool, make sure you push the audio into the loom after the node that generates it, or the export stage gets nothing. And yes, the Nodes 2.0 arrow-drag collapse quirk applies, with the pack's workaround in place.