Afloy Audio Info

If you've ever stared at a video workflow that comes out with drifting lip sync or a clip that ends half a second early, you know the urge to open a DAW just to ask "how long is this audio, actually?" Afloy Audio Info is the in-graph answer to that. You feed it one AUDIO wire and it hands back five numbers about the clip - no decoders, no ffmpeg, no model downloads, nothing to install beyond the pack itself.

It's part of the afloy_audio_tools pack, a small trio of audio utilities (trim, duration, and this info node) that all install together and show up under Add Node → audio → Afloy Audio Tools.

What it actually does

The node is pure Python introspection. ComfyUI's AUDIO type is a dict holding a waveform tensor and a sample_rate; this node reads waveform.shape and divides by the sample rate, then reports what it found. It doesn't decode or re-encode anything, so it's effectively free to run - and it tells you the numbers the pipeline actually sees, not the numbers printed on the filename. That's the useful part: a 48 kHz file that got loaded as 44.1 kHz will be reported as 44.1 kHz, and now you know why your sync math was off.

Feed it the audio input (only input there is) and you get:

• sample_rate (INT) - Hz, e.g. 44100
• channels (INT) - how many channels the tensor holds
• total_samples (INT) - raw sample count
• duration_sec (FLOAT) - total_samples ÷ sample_rate
• info (STRING) - a human-readable summary of all of the above

The info string is the sleeper hit. Wire it into a text display node and you can read the whole summary at a glance; or concat it into a filename if you're the kind of person who names outputs with their specs baked in.

How to install it

Same install as every node in this pack - it's one package that registers all three:

cd ComfyUI/custom_nodes
git clone https://github.com/afloy011-spec/afloy_audio_tools.git "Afloy Audio Tools"

Restart ComfyUI and you're done. ComfyUI Manager works too: Install Custom Nodes → search "Afloy Audio Tools". There's no requirements.txt - the node only uses numpy, Pillow, and torch, all of which ship with ComfyUI already. That's genuinely rare in this ecosystem, where most custom nodes drag in half of PyPI and then fight with each other over versions; this one just sits there.

When things look wrong

The most common "bug" here is actually upstream: if total_samples comes back 0 or duration_sec is 0, your Load Audio node fed you an empty clip - fix the file, not the node. If you expected stereo and it says 1 channel, that's the AUDIO tensor being mono, which is how many loads hand you the clip. And one tiny formatting note: the timecode-style strings in this pack are MM:SS with hundredths, not true milliseconds - fine for marking, just don't build precision-critical sync on the displayed digits.

It's not a flashy node. It's the "let's check our work" node, and in audio workflows that's often exactly the one you need.