Set Accumulator (audio)

Audio in ComfyUI gets the short end of the ecosystem's attention, and multi-branch audio comparisons are where that shows - you generate four versions of a clip and then have to wire every one by hand into whatever compares them. Set Accumulator (audio) is the audio member of the pack's accumulator family: you tag each branch's clip with a name and index, and a matching Get Accumulator (audio) gathers every Set that shares the name into a collected list.

How it works

Same boring-but-reliable pass-through as the other accumulators: the audio input comes in and leaves via the audio output unchanged. The mechanism is the label - name (default AUDIO) plus index, which auto-increments when you copy the node. A Get Accumulator (audio) with the same name collects every Set that shares it, in index order. The wiring is plain links created on the frontend, so execution and caching are standard - no hidden globals, which matters for audio nodes because they're already the fiddly corner of ComfyUI.

Inputs and output

• audio - the clip from this branch. Passes through untouched.
• name - accumulator name; must match the Get node's name exactly.
• index - position in the collected list, 0–9999.

The audio output is your pass-through, so the node can sit inline on the branch without breaking the flow.

Install

Part of the Kinburg-Nodes pack: ComfyUI Manager (search "Kinburg-Nodes") or git clone https://github.com/Kinburg/Kinburg-Nodes into custom_nodes, then restart. No extra dependencies for this node itself - but remember the pack installs llama-cpp-python automatically via install.py for its LLM nodes, which is unrelated to audio and harmless if you never use them.

The gotcha

Audio types are the least standardized socket type in ComfyUI - different packs expose different AUDIO structures, and a clip from one pack may not satisfy another's input. If the Get side collects clips that then fail to load downstream, the mismatch is usually format, not your wiring. And the standard accumulator rules apply: match the names exactly, and re-collect on the Get side after adding or removing Set nodes. This node is for collecting; the pack's Siren Compare and friends are the actual comparison side if you're doing serious audio A/B work.