PM Audio Loader
The node that needs ffmpeg on your PATH
- audio
Audio is the newest, thinnest layer of the ComfyUI stack - it only really showed up once video generation got good enough to want a soundtrack. So there aren't a hundred loaders to choose from, and PM Audio Loader is a decent one. It takes a file from your input folder and hands you a proper ComfyUI AUDIO output - a waveform tensor plus a sample rate - ready to feed into audio-conditioned video models, lip-sync chains, or whatever save/playback node you're running.
The "PM" branding here mostly means the same enhanced picker the pack applies to everything else: previews, an upload button, and a friendlier selection list than the bare core nodes. But the thing actually worth knowing about this node is its one sharp edge: it gets its audio by calling ffmpeg directly, and that has consequences.
How it works
Strip away the UI and the mechanism is refreshingly simple. The node shells out to ffmpeg with something like ffmpeg -i file -f f32le -, capturing raw float32 PCM audio from stdout. Then it greps ffmpeg's stderr output with a regex to find the sample rate and channel count (mono vs stereo), and packages it into the {"waveform": tensor, "sample_rate": int} dict that ComfyUI's audio nodes speak natively. If ffmpeg can't decode the file, you get a "Failed to extract audio" error with ffmpeg's own message attached, which is genuinely useful for figuring out what's wrong with the file.
Inputs and outputs
audio- a dropdown of audio files in yourComfyUI/inputfolder. That's the whole input surface.audio- the AUDIO output: a dict with the waveform tensor and sample rate, wired into any node that accepts AUDIO - audio-driven video conditioning, lip-sync, or a save node.
One input, one output. This is a loader in the purest sense.
Installing it
Same pack install as everything else here:
cd ComfyUI/custom_nodes
git clone https://github.com/pixel-magician/ComfyUI-PM-Manager.git
# restart ComfyUI
Or ComfyUI Manager, searching for ComfyUI-PM-Manager. The pack's own dependencies are opencv-python and imageio-ffmpeg, but here's the catch: this node does not use imageio-ffmpeg's bundled binary. It calls the literal ffmpeg command, so you need a system ffmpeg on your PATH. Windows users, that means actually adding it to PATH, not just having it somewhere on disk.
Common issues
The classic failure is FileNotFoundError: [Errno 2] No such file or directory: 'ffmpeg' or a "Failed to extract audio" error the moment you run it - both mean ffmpeg isn't reachable. Fix: install ffmpeg and make sure ffmpeg -version works in a terminal. After that, if a specific file fails, it's usually an unsupported or corrupt codec; the error text from ffmpeg will tell you. And the pack-wide rule: if the PM nodes don't show up at all, it's built on ComfyUI's newer extension API, so update ComfyUI first.
Inputs (1)
| Name | Type | Default | Description |
|---|---|---|---|
| audio | COMBO | 0 options: |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| audio | AUDIO | — |