MediaUtilities_AudioPreview

You've got an AUDIO wire in your graph and no idea what's actually on it. That's the whole job of MediaUtilities_AudioPreview: it's a terminal node that plays whatever audio you feed it, right in the ComfyUI interface, so you can check a clip before you render it out to disk. It's the audio equivalent of the built-in preview image node - nothing fancy, nothing clever, just a quick way to listen.

Why you'd reach for it

ComfyUI doesn't ship a core audio player in the graph, so once you start pulling audio in - from a URL, from a video you loaded, from one of the audio-generation or TTS packs - you're usually flying blind unless you save a file and open it yourself. This node closes that loop. Wire it up as the last stop on any AUDIO path and you get an instant sanity check: is this the right clip, right length, right vibe? The typical setup is Load Audio From URL (the pack's sibling node) straight into this, but it accepts AUDIO from anything that emits that type.

One thing to know before you fall in love with it: this is a preview node, not a passthrough. It takes audio in, shows you a player, and outputs nothing. It has to be the end of its branch of the graph. If you also want to save what you're hearing, run two nodes off the same AUDIO source - one into AudioPreview, one into Save Audio.

How it works

Under the hood it's short and honest. The node takes the AUDIO dict (which in ComfyUI is a waveform tensor plus a sample rate), writes it out as a FLAC file into ComfyUI's temp folder, and hands the UI the file's metadata so the frontend can render an audio player. That's the whole mechanism - there's no analysis, no waveform visualization, no clever audio processing. It saves to FLAC regardless of your source format, so don't read anything into the file type.

The one input that matters

There's a single required input, audio, and it expects the ComfyUI AUDIO type. That's it - no knobs, no settings. If you're staring at a missing-wire error, the fix is almost always that your upstream node outputs something else (an IMAGE, a tensor, a path string) rather than a proper AUDIO.

How to install it

MediaUtilities_AudioPreview lives in the ComfyUI-MediaUtilities pack. The easiest route is ComfyUI Manager: open it, search "ComfyUI-MediaUtilities", and install. Doing it by hand is just as simple:

cd ComfyUI/custom_nodes
git clone https://github.com/ThanaritKanjanametawatAU/ComfyUI-MediaURLLoader
cd ComfyUI-MediaURLLoader
pip install -r requirements.txt

Then restart ComfyUI. There are no model downloads - the pack is pure code. Its requirements (requests, numpy, pillow, torch, torchaudio, opencv-python) mostly overlap with what ComfyUI already ships; on most setups nothing new actually installs.

Heads-up on a README trap: the pack's own README tells you to git clone https://github.com/your-username/ComfyUI-MediaUtilities.git. That URL is a placeholder - it literally contains "your-username" and the repo is named ComfyUI-MediaURLLoader, not ComfyUI-MediaUtilities. Copying the README verbatim gets you a 404. Use the URL above.

Where people get stuck

If the preview comes back empty or nothing appears in the UI panel, the usual culprits: your audio wire isn't actually connected, or the upstream node failed and returned silence. This pack's audio loader (and the video loader's audio output) return an empty waveform on error instead of erroring the graph, so a "successful" run can be silent. Check the ComfyUI console for the Error loading audio line before you blame the player.

Also worth knowing: this is a small, lightly-maintained pack from a single author (last bumped to v1.1.1 in mid-2025). It works fine for what it does, but don't expect a big support community around it. For a preview node, that's a risk you can live with - the cost of failure is a missing player, not a corrupted model.