Audio Pad ⚡🅡🅞🅣🅘

The audio equivalent of adding margin to a page: this node extends a clip on either end by inserting extra samples, without changing anything about the audio that's already there.

Inputs and outputs

• audio - the clip to pad.
• pad_left and pad_right - how many samples to add on each side, 0 to 44100.
• pad_mode - what fills the padding: constant (silence - actual zeros), reflect (mirrors the audio back on itself at the edge), replicate (repeats the very edge sample), or circular (wraps around, pulling content from the other end of the clip).

Output is a single AUDIO.

Note the units: padding is specified in samples, not seconds, so the actual duration added depends on your sample rate - 44100 samples is one full second at a standard 44.1kHz rate, but far less time at a lower rate. That also explains the cap: 44100 samples is the field's max, which works out to about a second at standard rate, no more.

When you'd use it

Two common reasons: giving a clip a bit of silent headroom before a fade-in so the fade has somewhere to start from, or nudging a clip's length by a fraction of a second to match a frame count that's just barely longer than the source audio - avoiding an awkward gap at the tail of your render where the video keeps going after the audio has already ended.

For actual creative use, constant (silence) is what you want almost every time. reflect, replicate, and circular exist mostly for signal-processing reasons - avoiding hard edge artifacts in some downstream analysis step - rather than sounding natural on their own. Don't expect them to sound like a real continuation of the clip; they're technical padding modes, not a creative fill.

Installing it

ComfyUI Manager: search RyanOnTheInside, install, restart. Manual:

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

No model downloads needed.

Where people get burned

The 44100-sample cap per side is fixed in the node - if you need more than about a second of padding, you'll need to chain two AudioPad nodes back to back rather than expecting a single call to cover it. And if you reach for reflect or circular expecting it to sound like a natural extension of the music, it usually won't - for anything meant to be heard rather than just processed, stick with constant silence and handle the actual audio content some other way (repeating a section, generating more, etc.).