Empty Audio

EmptyAudio generates exactly what the name promises: silence. A blank AUDIO tensor of a chosen duration, sample rate, and channel count. It sounds like a node that exists to fill a dropdown, but it's quietly one of the more useful tools in the audio family - because generative audio is almost never the right length, and silence is how you pad things out to match.

Think about the workflow problems that are really timing problems. Your video is 20 seconds but the generated track is 15 - you need 5 seconds of quiet tail so the audio lines up when you mux them. Or you're assembling a fixed-length piece with AudioConcat and need to insert a pause between sections. Or you want a placeholder track so you can build and test a pipeline before you have any real audio to feed it. All of those are EmptyAudio's job.

How it works

It's a pure constructor: it allocates a zeroed float32 tensor of duration × sample_rate samples with the requested number of channels, and labels it with the sample rate you chose. No processing, no hidden math - a block of silence in the exact shape you asked for.

The three inputs, all of which you'll actually set:

• duration (FLOAT, default 60) - length in seconds. This is the one you'll change constantly.
• sample_rate (INT, default 44100) - samples per second, from 1 to 192000. The default is CD standard; use 48000 if you're targeting video (that's the video convention) so the silence matches your other sources.
• channels (INT, default 2) - 1 for mono, 2 for stereo. Match this to whatever you're padding against.

The output is a standard AUDIO tensor, so it plugs into anything the other audio nodes produce: concatenate it, merge it, save it, preview it. It'll be silent on playback, obviously - which is a great way to verify your routing is actually wired correctly before you add real audio.

Inputs and output

• duration (FLOAT, default 60) - length of the silence in seconds.
• sample_rate (INT, default 44100) - sample rate, 1–192000.
• channels (INT, default 2) - 1 or 2.
• AUDIO (output) - the silent clip.

Getting it

Ships with ComfyUI core since September 2025. Nothing to install.

Common issues

The only real pitfalls are the ones you'd expect: a sample rate or channel count that doesn't match your other tracks. Downstream nodes handle mismatches by resampling and duplicating mono to stereo, so it won't error - but if you're padding a video's audio track, just set the sample rate to 48000 and channels to 2 from the start and you'll avoid the silent re-encode. And remember: silence padded onto a track still counts as signal for length-matching purposes, which is exactly why this node exists.