How to split an audio file

Splitting one long session into many separate files, one per item, is a different job from cutting a single piece out of the middle of a recording.

One-into-many is the case for a batch of voiceover lines, an album from one continuous recording, a set of chapters, or a session where you recorded twelve takes back to back. That is what this page is mostly about, and it can be almost entirely automated.

Cutting one section out is ordinary editing: select, delete, export. If that is what you need, any editor does it in a minute.

The third case is splitting a file only because it is too large to send. There, a compressed export at a lower bitrate is usually a better answer than several pieces someone has to reassemble.

For a handful of splits, a browser-based splitter is the fastest path: upload, drag markers to the split points, download the pieces. Nothing to install and nothing to learn.

The thing to check is whether the tool re-encodes. Splitting an MP3 and re-encoding it loses a little quality each time, which matters if the file will be edited further.

Hilite has a split audio tool that does this in the browser, and we make it, so weigh that as you like. Any similar tool works the same way.

For anything sensitive, prefer a tool that processes in the browser rather than uploading to a server, or use desktop software.

Open the file in Audacity. Place the cursor at the start of each section and press Control and B to add a label, typing a name for each one. Those names become the filenames.

Then File, Export, Export Multiple, and choose to split based on labels. Audacity writes one file per label in a single pass.

Twelve files takes about two minutes this way, and the same amount of manual selecting, exporting and naming takes most of an hour.

Name the labels as you place them. Audacity will happily produce twelve files called untitled, and renaming afterwards is worse than typing at the time.

If your recording has clear gaps between items, you do not need to place labels by hand at all. Audacity's Analyze menu has Silence Finder and Label Sounds, which place labels automatically wherever the audio drops below a threshold for long enough.

Set the minimum silence duration to something longer than your natural pauses, a second or two, or it will split in the middle of sentences.

Check the labels it produced before exporting, and delete or move any that landed wrong. It is much faster to correct a few automatic labels than to place forty by hand.

This is the route worth knowing about if you record many short items in one sitting. Leave three seconds of deliberate silence between takes and the splitting becomes automatic.

Every time an MP3 is decoded and re-encoded it loses a little. One round is usually inaudible on speech. Several rounds accumulate, so if you will process the pieces further, split to WAV and compress only at the end.

Cut in silence wherever possible. A split that lands mid-word gives you two files that both sound broken, and a split during room tone is invisible.

Leave a little silence at the start and end of each piece rather than cutting tight to the first and last word. A file that begins the instant a word starts sounds clipped even when nothing is missing.

If a split clicks, you cut mid-waveform. In Audacity, press Z before splitting to snap the selection to a zero crossing.