🔊✂️AGSoft Audio Split Plus

🔊✂️AGSoft Audio Split Plus is what you grab when a plain cut isn't enough. It does everything the regular Audio Split does - cut a fragment by start_time / end_time, with fast (stream copy) or precise (re-encode) modes - and then layers on the stuff that turns a raw cut into something usable: fade in/out that kill the click at the cut points, loudness normalization, reverse, and loop.

The practical use case that sells it: cutting a tight loop out of a music track or a TTS phrase, de-clicking the edges so the loop doesn't pop, and repeating it N times. That's three nodes' worth of work in most packs, and here it's one node.

What's new over the plain split

• fade_in / fade_out (0–60s) - ramps at the start and end of the fragment. The tooltip's honest about why: they remove the clicks you'd otherwise hear at hard cut points. 0 = off.
• loudnorm - normalizes the fragment to EBU R128 (integrated −16 LUFS). Useful when you're slicing samples out of different sources and want them to sit at the same level.
• reverse - plays the fragment backwards. One flag, instant psychedelia.
• loop (1–100) - repeats the fragment N times. 1 = no looping.

One mechanism detail worth knowing: when any effect is on, re-encoding is unavoidable - you can't stream-copy a reversed fragment. If format is source in that situation, the node re-encodes with the codec that matches the source extension, so you keep the same file type rather than defaulting to something random. Loop is also done as a two-pass operation: cut to a lossless temp file first, then loop and encode - which avoids the quality spiral you'd get looping a lossy encode repeatedly.

The same source gotcha, still true

Everything from the plain version carries over: if audio comes in through the audio AUDIO socket (rather than audio_path), it's already decoded PCM, so format=source yields a WAV - not your original MP3. And stream copy only really happens when the file is connected via audio_path and no effects are used. Any effect flips you into a re-encode, so don't expect the fast mode to stay lossless once you touch the fades.

Outputs

Same generous set as the plain split: audio_path (STRING, where it landed), audio (AUDIO, for chaining), plus duration_seconds, duration_timecode, file_size_mb, sample_rate, and channels.

Install

Part of the comfyui-AGSoft pack. ComfyUI Manager → search "comfyui-AGSoft" → Install → restart, or by hand:

cd ComfyUI/custom_nodes
git clone https://github.com/Art-xmaster/comfyui-AGSoft

Restart and go. No models to download; it drives the bundled FFmpeg binary (or a system ffmpeg on PATH).

If you only ever need a clean, lossless, instant cut - no effects - the plain Audio Split is the lighter choice. The moment you want edges that don't pop, reversed audio, or a seamless loop, Plus earns its name.