ITL Multi Audio Loader

ComfyUI has audio support now - a real AUDIO type that video and audio-generation models actually consume - but the stock loading story is still "one file per node." When you're building a workflow that takes several voice clips, SFX samples, or music stems and routes each one somewhere different, the ITL Multi Audio Loader is the same "one node, many files, many sockets" pattern as its image sibling: drop up to 8 audio files on it, and each gets its own audio_N output.

How it works

Drop files onto the field (or click to browse), and they upload into ComfyUI's input/ folder like any other loader. Each row gets a socket - audio_1, audio_2, and so on - in row order, and each row shows a duration so you can tell your clips apart at a glance. Drag the ⠿ grip to reorder, ✕ to remove, or Sort by name for folder order.

The one audio-specific behavior worth knowing: the node never resamples. Each audio_N output keeps its file's native sample rate. That's deliberate - if a downstream node expects a specific rate, resampling is a separate, explicit step, not something this loader silently does for you. Some loaders quietly convert and you get muddled output; this one hands you exactly what's in the file.

The settings and the trap

• output_slots - auto makes sockets appear/disappear with the file list. Pin a fixed count (1–8) and the sockets stay put while you swap files, keeping your wires intact. Extra pinned sockets with no file behind them output nothing.
• The row on/off toggle - switching a row off keeps its socket position (nothing shifts) but that socket outputs None on the next run. The node equates this with leaving an unplugged OPTIONAL input, which is safe - unless that socket feeds a REQUIRED input downstream, in which case the node fails on None. Don't toggle off a row that feeds something mandatory.
• count - the INT output counts enabled files actually emitted, not rows. Toggle a row off and count drops.

Reordering or removing files shifts what each socket carries, and the node warns in its status line whenever a change touches a socket that has a wire. Trust that warning - the repo ships with issues disabled, so there's no support ticket to file if you miss it.

Outputs & install

Outputs: count (INT) plus audio_1 through audio_8 (AUDIO). The Advanced variant adds filename_N per file, which is handy when you want to know which file fed which socket after the fact.

Install through ComfyUI Manager (search "ComfyUI-IntoTheLatent-Utils") or:

cd ComfyUI/custom_nodes
git clone https://github.com/Into-The-Latent/ComfyUI-IntoTheLatent-Utils

then restart. No extra dependencies - the pack's requirements are just Pillow, which ComfyUI already ships, and audio decoding rides on ComfyUI's own av support.

This node is simple, and that's the whole appeal: if you've ever stacked six Load Audio nodes to feed a multi-speaker workflow, this replaces them with one drop target and no socket-numbering mistakes.