获取音频地址

AutioPath (获取音频地址, "get audio path") is the pack's audio version of ComfyUI's built-in Load Image: a file picker that returns a filesystem path instead of the file itself. Pick an audio file from your input folder, and out comes its absolute STRING path, ready to hand to anything that wants a path rather than an AUDIO tensor.

That last sentence is the reason this node exists. ComfyUI's core audio loading mostly speaks in AUDIO tensors, but a whole layer of utilities - ffprobe-based info nodes, pydub merging, TTS tools, anything shelling out to ffmpeg - wants a real path on disk. AutioPath is the bridge. It slots neatly in front of this pack's AutioInfo (reads metadata via ffprobe) and MultiMergeAudio (concatenates files with pydub), both of which take paths, not waveforms.

How it works

Under the hood it's the standard ComfyUI input-file widget: it lists audio files in ComfyUI/input, with the upload button (audio_upload) enabled so you can drop in a file from the browser. The audioUI input is the AUDIO_UI widget ComfyUI uses for the audio-picker chrome - you mostly ignore it. The node resolves the selected file to its full annotated path and returns it.

One detail that makes it trustworthy in loops: it implements IS_CHANGED by file modification time, so if you swap or re-upload the audio while a workflow is cached, the node notices and re-runs. That's the kind of thing stock path nodes often forget and it bites people in batch runs.

Inputs and output

• audio - dropdown of audio files in your input directory (upload supported).
• Output: 音频地址 (audio path), a STRING - feed it into path-consuming nodes like AutioInfo, MultiMergeAudio, or external ffmpeg tooling.

The elephant: the typo

The class in the source is literally named AudioPath, but it's registered as AutioPath and that's the spelling you'll see in the node list and in workflow JSON. It's a typo baked into the pack, and nobody's fixed it because the pack's author named its sibling AutioInfo the same way. Don't fight it; search "Autio" in the node picker and you'll find both.

Install

Part of yanlang0123/ComfyUI_Lam - Manager search "ComfyUI_Lam", or:

cd ComfyUI/custom_nodes
git clone https://github.com/yanlang0123/ComfyUI_Lam

restart, and you're done. No models, no keys. This one loads with zero of the pack's heavy dependencies, so there's no reason to run the README's install.bat unless you're using the audio/TTS machinery that genuinely needs pydub and librosa. If you do use MultiMergeAudio downstream, make sure ffmpeg is reachable - pydub depends on it for MP3 handling.