Save Audio 🔊

This is the small, boring node that actually gets your sound out of ComfyUI, and you need it more often than you'd think. Most audio nodes - including the Edge TTS node it ships beside - produce an AUDIO object in memory and stop there. Save Audio is the one that takes that object and writes a real file to disk in the format you want. Nothing glamorous. Just the piece that turns "the graph ran" into "there's an mp3 I can use."

It comes from 1038lab's ComfyUI-EdgeTTS pack, and it's the natural terminal node for the Edge TTS → Save Audio flow, but it'll happily save AUDIO from anything upstream - a TTS node, a loaded clip, an audio-processing chain.

How it works

You hand it an AUDIO input and it encodes and writes the file. It's a true output node, so dropping it into your graph and hitting Run is enough to produce a file - no separate "execute" step. It picks the encoder from your chosen format and quality preset, handles the file naming, and - unless you tell it otherwise - auto-numbers the output so a second run doesn't clobber the first.

Handy detail: it doesn't just save, it also passes the audio straight back out. So you can drop it mid-chain to write an intermediate file and still keep feeding the same audio into the next node.

The inputs and outputs that matter

• audio (AUDIO) - the sound to save. Wire this from Edge TTS or any node that emits AUDIO.
• format - wav, mp3, or flac (default mp3). WAV for a lossless, universally-openable file; MP3 when you want something small to hand off; FLAC for lossless-but-compressed.
• quality - high / medium / low (default high). For lossy MP3 this sets the bitrate; on WAV/FLAC it makes little practical difference, so just leave it on high.
• filepath - where and what to name it. Leave it blank to drop the file in ComfyUI's default output folder with automatic numbering; set a path/name to control it.
• overwrite (default false) - off means it numbers new files instead of overwriting; flip it on if you deliberately want to replace the same file each run.

Outputs are filepath (a STRING with the path it actually wrote - useful if a later node needs to know where the file landed) and audio (the same AUDIO, passed through).

How to install it

It's part of the ComfyUI-EdgeTTS pack, so you install the whole pack once. Via ComfyUI Manager: search Comfyui-EdgeTTS, install, restart. Or manually:

cd ComfyUI/custom_nodes
git clone https://github.com/1038lab/ComfyUI-EdgeTTS.git
./ComfyUI/python_embeded/python -m pip install -r requirements.txt

Then restart ComfyUI.

Common issues

WAV saves fine but MP3 or FLAC fails or produces nothing. Lossy/compressed encoding leans on the pack's audio backend. If you're on a recent torch/torchaudio (2.9+), that stack now needs torchcodec to encode - the pack's own install notes call this out. The fix is to match torchcodec to your torch build:

./ComfyUI/python_embeded/python -m pip install --no-cache-dir "torchcodec==0.9"

FFmpeg on your system PATH is the other thing the pack expects for audio work; if MP3 export is misbehaving, confirm that's installed too.

"It ran but I can't find the file." With filepath blank it writes into ComfyUI's default output directory - check there. The filepath STRING output tells you the exact path if you ever lose track.

Every run makes a new numbered file. That's overwrite doing its job (it defaults to off). Turn it on if you'd rather keep overwriting one filename.