ADV Audio Series (4)

Audio nodes are rarer in ComfyUI than image nodes, and picking between several generated clips by hand is usually a "save them all, listen, reconnect" chore. ADV Audio Series (4) exists to make that chore a one-click decision: it takes up to four AUDIO inputs and emits them as a list, ready for the ADV Pick Any Chooser from the same pack.

The payoff is a dialog where each candidate audio shows up with a play button, so you actually listen before you commit. That's the pattern this whole pmarmotte2/Comfyui_Pick_Any pack is built around - generate several options, pick the winner interactively, carry it onward.

How it works

In ComfyUI, an AUDIO value is a dict with a waveform tensor and a sample_rate - that's the convention shared by the stock audio nodes. What this node does is two things at once:

1. It flattens each of its four slots into individual audio entries, splitting any batched waveform tensor so each clip is its own selectable item rather than one combined tensor.
2. It bundles them into a single list and emits it as the audios output.

Anything None (an unconnected slot) is dropped on the way through, so partial wiring just works. When the chooser receives the list, it converts each entry to WAV on the fly for the in-dialog playback - no extra conversion steps on your side.

Inputs and outputs

• audio_1 through audio_4 - optional AUDIO inputs. Slots you leave empty are skipped.
• audios - the output, a list of audio dicts, wired into the chooser's audios port.

Install

It's part of the Comfyui_Pick_Any pack, so install the pack once and you get every node in it:

cd ComfyUI/custom_nodes
git clone https://github.com/pmarmotte2/Comfyui_Pick_Any

Restart ComfyUI and hard-refresh the browser tab. There's no requirements.txt and no model download - pure utility, nothing to break your environment. Keep the folder name free of spaces, per the README.

Common issues

• "The chooser shows no audio." If your source node hands over an empty batch, there's nothing to split, and the node can't conjure clips out of nothing. Generate first, then bundle.
• "My audio came out as one giant item instead of several." That's actually the intended behavior when the input is already a single batched tensor - the node splits it into individual entries so each is selectable. If your source emits several separate dicts instead, they each pass through as their own item. Either way you end up with per-clip options.
• Need eight slots? The pack also ships ADV Audio Series with eight inputs. Same node, more room.