Load Audio

Audio is the thing nobody's workflow includes until they need it, and then they need it badly. You're putting a soundtrack under a generated clip, driving visuals with an audio-reactive graph, or muxing an LTX-2 video that came out of the model without sound (that's the standard complaint - LTX-2 generates synchronized audio natively, but if yours arrived mute, you're adding a track by hand). YogurtLoadAudio is the front door: pick a file, optionally trim it, get an AUDIO output.

It's part of ComfyUI-YogurtNodes, yogurt7771's 150+ node all-in-one. The pack is obscure - MIT-licensed, auto-generated README, zero community footprint - and this node is one of its IO corner staples: a plain, no-nonsense audio loader with the same start/end trimming as the video loader.

How it works

The audio input is a dropdown populated at runtime from your ComfyUI input folder - drop my_track.mp3 into ComfyUI/input and it appears. There's also an audioUI input, which is the file-upload widget you get in the node UI. Then two time fields, both in seconds: start_time (0 = from the beginning) and end_time.

The gotcha is in end_time. 0 means "no end limit", not "zero seconds". Leave it at 0 to load the whole file, or set it to e.g. 15.0 to cut everything after 15 seconds. Both fields step by 0.01s, so you can trim fairly precisely.

Output is a single AUDIO - the standard ComfyUI audio format, so it wires into whatever audio-aware nodes you're already using: video-combine nodes that accept an audio track, audio-reactive processing, or anything downstream that expects AUDIO.

Where you'll use it

• Adding a music or effects track to generated video before the final combine.
• Trimming a long file down to the section you actually need, so downstream nodes don't chew on the whole thing.
• Audio-reactive workflows where the waveform drives masks or timing.

The one thing to remember is the folder rule: it has to live in ComfyUI/input to show up in the dropdown. If your audio sits elsewhere on disk, the pack's sibling YogurtLoadAudioPath takes an absolute path instead - same trimming, no file-copying dance.

Install

ComfyUI Manager, search ComfyUI-YogurtNodes, install, restart. Or:

cd ComfyUI/custom_nodes
git clone https://github.com/yogurt7771/ComfyUI-YogurtNodes
cd ComfyUI-YogurtNodes
pip install -r requirements.txt

Restart, look under "Yogurt Nodes". Dependencies are the pack's usual light set (numpy, pillow, requests, opencv-python) - the heavy stuff like google-genai only loads if you touch the LLM nodes, so nothing here drags in extra weight or model downloads.