Load Audio

Kling's audio-producing nodes - custom voice cloning, TTS, video-to-audio - hand you URLs to sound files. Preview Audio is the node that takes those URLs and makes them real: it downloads the audio, saves it to your output folder, and loads it back into ComfyUI as an AUDIO connection you can preview, keep, or feed into another node. It's the audio twin of Preview Video, and it's the missing piece of the pack's voice workflows.

You'll recognize it in the node list under a slightly different name than the others - on comfy.icu it's indexed as "Load Audio" (class LoadAudio) even though the pack's README calls it Preview Audio. Same node, one naming hiccup. What matters is the job: audio URL in, playable file out.

How it works

The main input is audio_url - a string URL. The README's canonical use is delightfully direct:

• Connect Custom Voice Create.trial_url (or Custom Voice Query.trial_url) straight into audio_url.

That's the audition loop: clone a voice, get its trial preview URL, and hear the clone right in ComfyUI before you commit credits to a full generation with it. You can also point it at any other public audio URL the pack produces - a TTS result, a video-to-audio render, whatever.

On the node you'll find a filename prefix and a save toggle, mirroring Preview Video: save to your output directory as a proper audio file, or just preview without writing to disk.

Outputs

The node returns the audio twice over, which is genuinely handy:

• AUDIO - the audio as a ComfyUI AUDIO object. This is the one that matters: it means you can chain the downloaded audio straight into another node that accepts AUDIO, like Lip Sync Audio Input or Custom Voice Create.
• file_path - the local path to the saved file, for anything that needs a path instead.

So the audition flow doubles as a file-download flow: trial URL in, a saved audio file plus a usable AUDIO connection out.

Install and gotchas

Standard pack install: ComfyUI Manager (search "Comfyui-Kling-Wrapper") or git clone https://github.com/magicwang1111/Comfyui-Kling-Wrapper.git into custom_nodes, pip install -r requirements.txt, restart, Kling API key in config.local.json. No models, no metered cost beyond the download - previewing is the cheap part of any voice workflow.

The gotcha is the same class as Preview Video: it fetches whatever URL it's given, so expired or failed URLs produce an error rather than a file. And if you're expecting to see "Load Audio" as a file-picker for local files - the class name suggests it - that's not this node. This is a URL-driven downloader/previewer; for local files you'd use ComfyUI's own audio loading nodes. For everything this pack generates as audio, it's the natural endpoint of the wire.