Load & Trim Audio

If you've ever loaded audio into ComfyUI just to trim five seconds off the front, you know the drill: open an external editor, cut, export, re-import. Load & Trim Audio kills that whole detour. It's an audio loader with a full interactive waveform drawn right on the node - click to place a playhead, drag the shaded region to select, hit play to preview - and it hands you back the sample-accurate trimmed AUDIO plus its duration. No ffmpeg one-liner, no typing sample numbers.

It's one node in Enigmatic Nodes, a small utility pack by the individual developer enigmatice. Pure Python plus a bit of JS; no models, no API keys, nothing to download. The kind of node that's easy to overlook until you've wired audio into a video pipeline and realized half the work is just "get the clip to start where I want."

How it works

The backend is deliberately boring, and that's a compliment. It decodes your file through ComfyUI core's own PyAV-based audio loader (comfy_extras.nodes_audio) and trims by sample count, so the cut is exact rather than approximate. One genuinely thoughtful detail: the code explicitly avoids torchaudio.load because on torchaudio 2.9+ that routes through the optional torchcodec backend, which isn't installed in many portable builds and just raises ImportError. The author worked around a real footgun instead of shipping it.

The fun lives in the frontend. web/audio_trim.js draws the waveform on the node: click to drop the playhead, press I/O to set in/out points there, drag the shaded region's edges to trim by hand, drag the middle to slide the whole selection. ← / → nudge the playhead, Shift makes the step bigger, Space (or Play) previews just the trimmed range, and double-click resets to the full clip. Once you've felt how fast that is compared to typing start_time: 1.437, the old way feels like filing paperwork.

The inputs that matter

• audio - a dropdown of files in ComfyUI's input folder, with an Upload button on the node (or drag a file straight onto it). It also accepts video files and pulls the audio track, which is handy when your source is a clip.
• start_time / end_time - seconds into the clip. end_time of 0 (or anything past the end) means "play to the end", which is why the default does the right thing the moment you add the node.

The outputs

• audio - a ComfyUI AUDIO object (waveform tensor plus sample rate), ready to feed audio nodes or audio-driven video workflows.
• duration - the trimmed length in seconds as a FLOAT, useful for wiring into other math or metadata downstream.

Installing it

This is the same for every node in the pack:

• ComfyUI Manager - search Enigmatic Nodes, install, restart.
• Manual - cd ComfyUI/custom_nodes && git clone https://github.com/enigmatice/comfyui-enigmatic-nodes, then restart.

No model files to fetch. The pack's pyproject.toml declares torch, numpy, scipy, Pillow, and opencv-python. ComfyUI core already ships all of those except opencv-python - and you'll almost certainly have that already, since it's one of the most common transitive dependencies across ComfyUI node packs. Manager installs dependencies for you; if you cloned by hand and something complains, pip install opencv-python covers it.

Common issues

• Your file isn't in the dropdown. It has to live in ComfyUI/input or be uploaded through the node's button. Drop files elsewhere and they won't appear.
• end_time seems to do nothing. Remember 0 means "to the end", so don't use it hoping for an empty clip.
• The waveform looks empty. The node draws it from the loaded file, so check you picked something with an actual audio stream.

It's a young pack - this node shipped mid-2026, with zero community footprint so far - so don't expect a stack of tutorials around it. But for the narrow job of "trim audio without leaving the graph," it does the thing exactly the way you'd want it done.