Resample Audio

ResampleAudio changes the sample rate of an AUDIO track, and if you've spent any time with this pack you'll quickly see it as the escape hatch. ConcatAudio and JoinAudio both refuse to combine clips with different sample rates, and those errors are the single most common thing beginners hit. This node is the fix. It ships in lebrosoft/ComfyUI-VideoChatWrapper, a small pack whose audio utilities are the actually-useful half of the repo.

How it works

It's a thin wrapper around torchaudio.transforms.Resample, which is to say it's doing proper, high-quality sinc interpolation rather than a cheap nearest-neighbor hack. The defaults are sensible, so most of the time you just set a target frequency and run. When the two tracks in a ConcatAudio are 44.1 kHz and 48 kHz, you run the 44.1 kHz one through this node with new_freq = 48000 and the join stops complaining.

The inputs that matter

There are more knobs here than the other audio nodes, but realistically you touch two:

• audio - the track to resample.
• new_freq - the target sample rate in Hz. Default is 32000, which is a strong hint this pack was written with speech and ASR pipelines in mind (32 kHz is a common speech-model rate). If you're mixing music or general audio, you probably want 44100 or 48000.

The rest are advanced interpolation settings you can safely ignore: resampling_method (sinc_interp_hann vs sinc_interp_kaiser), lowpass_filter_width (default 6), rolloff (default 0.99), and an optional beta for the Kaiser window. The defaults are the standard, well-tested torchaudio values. Tweak beta only if you know what a Kaiser window is and why you'd want a different shape.

Output is a single audio track at the new sample rate.

How it fits the workflow

Think of it as the normalization step. The pattern for joining clips in this pack is: ResampleAudio to line up rates → ConvertAudioChannels to line up mono/stereo → ConcatAudio or JoinAudio to actually splice. Once you internalize that three-node chain, the pack's audio half stops being a source of mysterious errors and becomes genuinely useful.

Installing it

Same as every node in the pack. ComfyUI Manager, search ComfyUI-VideoChatWrapper, install. Or manually:

cd ComfyUI/custom_nodes
git clone https://github.com/lebrosoft/ComfyUI-VideoChatWrapper

Restart ComfyUI. ResampleAudio runs on torchaudio, which is bundled with the portable ComfyUI install - if you're on a custom environment and get a ModuleNotFoundError, pip install torchaudio fixes it. Nothing else in the pack's dependencies applies here; requirements.txt only lists qwen_vl_utils for the video-chat nodes, and no model downloads happen for audio work.

Fair warning: this pack has essentially zero community presence, so there's no stack of forum threads about it. It's not a scam and not broken - just quiet. For resampling, the node is a thin, correct wrapper around a battle-tested library. That's exactly what you want a utility node to be.