How Auto-Cut Works in AudioMultiCut and How to Get Clean Results

Auto-Cut analyzes the waveform locally in your browser and looks for quiet regions long enough to be treated as real breaks. It then groups the loud regions between those breaks into segments, applies a minimum clip length so it does not produce dozens of tiny snippets, and shows the results as segment cards under the waveform.

Three numbers do most of the work: how long a quiet region has to be before Auto-Cut treats it as a real split, how long a clip is allowed to be at minimum, and how much short silence is allowed to stay inside a clip without breaking it. Each preset is a different combination of those values tuned for a specific kind of recording.

The preset row is the most important choice. Picking a matching preset usually gets you 80 to 90 percent of the way there before you adjust anything by hand.

Smart is a balanced default for mixed recordings. Rehearsal Songs is tuned for band practice where you want full songs but not count-ins. Live Show is more forgiving inside a track and stricter about what becomes a new song, so applause does not turn into its own clip. Lecture Chapters waits for a real section break and ignores normal speaking pauses. Podcast Clips aims for longer beats instead of every short pause. Lesson Clips keeps short explanation gaps inside the clip but breaks at longer teaching resets. Samples is the only preset built for very short repeated sounds with small gaps, like one-shots across keys.

If the preset is close but not quite right, open Manual override under the preset row. The four sliders map directly to the underlying detection parameters: Window controls the analysis resolution, Min clip is the minimum length any segment is allowed to be, Keep inside controls how much short silence can stay inside a clip without breaking it, and Split after is the silence length that finally counts as a real split.

The most useful slider in practice is Split after. If Auto-Cut is splitting too eagerly, increase it. If it is gluing two clips together that should be separate, reduce it. Move it in small steps and re-run. Min clip is the second most useful: raise it if you keep getting tiny fragments, lower it for sample-style recordings.

Auto-Cut gets you to a starting list of clips, but a few small touches almost always make the export feel cleaner. The waveform preview, the focused playback per segment, and the boundary editor are designed for exactly this last pass.

Spending a minute on these touches turns a useful first pass into a set of clips you would actually send to someone else.

Auto-Cut is silence-based, not content-based. It does not transcribe speech, identify speakers, or detect chord changes inside a continuous performance. It also does not try to remove silence inside a clip; it only uses long silences as boundaries.

That is intentional. The job is to give you a fast, reliable first pass that runs entirely in the browser without uploading your audio anywhere. The detailed creative decisions about what each clip should be are still yours, which is the part you usually want to keep.