Save Audio With Path (CRT)

Stock ComfyUI save nodes are opinionated about where your files go - everything lands in the same output directory with a shared naming convention, which is fine until you're running a project with a lot of audio output and want it organized somewhere sensible. This node is a straightforward upgrade: save your generated audio as an uncompressed WAV, but to whatever folder and subfolder you actually want, with real control over bit depth and what happens if your audio is clipping.

The inputs that matter

Most of these have tooltips written well enough that they don't need much interpretation, but a few are worth calling out:

• audio - your generated or processed audio, from a TTS node, a voice model, a music generator, whatever's upstream.
• folder_path (default /tmp/ComfyUI/output) and subfolder_name (default audio) - the actual point of this node over the stock one. You get a real, arbitrary path rather than being locked to ComfyUI's default output tree. Worth flagging: that default path looks like it was set on a Linux dev machine - if you're on Windows, don't trust the default blindly, set it explicitly.
• filename and suffix - base name and an optional tag appended to it.
• sample_rate (44100 / 48000 / 88200 / 96000 / 192000) - a fallback used only if the incoming audio data doesn't already specify one.
• bit_depth (24-bit PCM / 32-bit float) - real audio-engineering control most save nodes don't expose.
• overwrite (default true) - off means a numbered suffix gets appended instead of clobbering an existing file.
• normalize_clipping (default true) - the one setting worth actually understanding rather than leaving on autopilot. Its own tooltip lays out the tradeoff precisely: enabled, it peak-normalizes audio that's clipping, turning the whole file down just enough to fit cleanly. Disabled, 24-bit PCM will hard-clip anything over 0dB at -1dBFS (audibly harsh distortion), while 32-bit float will happily save samples over full scale without clipping them at all, at the cost of a file that isn't standard playback-ready until something downstream brings the level back down.

The output that matters

There isn't one - this is an is_output_node, meaning it's a terminal node in your graph. It writes the file and that's the whole job.

How to install it

Part of CRT-Nodes, which - a little unusually for a pack this heavy on image and video FX - ships a genuine small audio post-production suite alongside everything else: a parametric EQ, a tube compressor, mono-to-stereo conversion, an audio transcript node. This save node is the natural endpoint of that chain. Via ComfyUI Manager, search CRT-Nodes and install. Manually: git clone https://github.com/PGCRT/CRT-Nodes.git into custom_nodes, pip install -r requirements.txt, restart. Note that pedalboard and librosa - both real audio-processing libraries - are in the pack's base requirements, so the audio tooling isn't an afterthought bolted onto an image pack; it's built with the right dependencies from the start.

Common issues & troubleshooting

Audio sounds harshly distorted at the loudest moments. That's 24-bit PCM hard-clipping - either turn normalize_clipping on to have the node handle it automatically, or fix the gain staging upstream before it reaches this node.

File isn't where you expected it. Check folder_path first - the default is a Linux-style temp path, and it's easy to leave it unchanged and then go hunting through ComfyUI's usual output folder instead.

Existing files keep getting silently replaced. overwrite defaults to true. Turn it off if you want each run to land as a new, numbered file instead.