Split a long recording into pieces.

At a glance

Which mode to use

Equal parts divides the file into a fixed number of pieces of identical length. This is the one to use when the constraint is a count — you need the recording in four chunks because that is how many you are sending, or because a platform caps uploads at a certain size and dividing by four gets you under it.

Fixed length cuts every N minutes regardless of how many pieces that produces. This is right for chapterising: a two-hour lecture in ten-minute segments, an audiobook in half-hour chunks that match how long someone listens in one sitting.

Silent gaps finds the pauses and cuts there. This is the mode for material with natural boundaries — an album side recorded in one pass, a set of takes with a break between each, a recording of several songs. It gives you pieces that start and end where the content does, rather than slicing through the middle of a word.

When you'd want this

• Chapterising an audiobook or lecture so it can be listened to in sittings, and so a player remembers where you were.
• Getting a long recording under an upload limit — a two-hour interview that will not fit into a transcription service or an email in one piece.
• Splitting a vinyl or tape transfer into tracks. Recording a whole side in one pass and cutting at the gaps is far quicker than starting and stopping.
• Separating takes from a rehearsal or practice session recorded continuously.
• Cutting a long file into transcription-sized pieces, then prepping each one for whichever service you use.

Getting silence detection right

Two settings control it. Silence below is how quiet a passage has to be to count as a gap. −45 dB works for most recordings. If your recording has audible room tone the threshold needs to be higher (−35 dB) or nothing will register as silent; if it was made in a treated room, −55 dB avoids cutting at every small pause.

Gap at least is the minimum length of quiet before it counts as a boundary. This is the setting that stops it cutting between every sentence. Two seconds is right for music with gaps between tracks; three to five seconds is safer for speech, where natural pauses are common and you only want the deliberate breaks.

If you get too many pieces, raise the minimum gap. If you get none at all, the tool will say so rather than silently returning the whole file — lower the threshold or raise it depending on whether the recording is noisier or quieter than you assumed. Pieces shorter than a second are discarded automatically, since those are almost always an artefact rather than content.

What you lose

A generation of quality, if you output to MP3 from an MP3. The audio is decoded and re-encoded, so use WAV if the pieces are going into an editor, or 320 kbps if they are going out into the world. This is the same trade as any lossy conversion on the site.

Metadata. The pieces come out with sequential filenames and no tags, so an audiobook split into chapters will need titling — the tag editor handles that without re-encoding anything.

Precision, in silence mode. Cuts land in the middle of a detected gap, which is usually what you want, but it means the boundary is approximate rather than sample-accurate. For an exact cut point, the cutter gives you a waveform and draggable handles instead.