Split Audio into Parts

One recording in, many files out — an album into its tracks, a lecture into chunks, an audiobook into its chapters, a long interview into pieces a transcription service will accept. Five ways to decide where the cuts go, and a split that changes nothing about the sound produces files that are byte-for-byte the original.

• Five ways to place the cuts: an equal number of pieces, a fixed length, marks you place yourself, gaps found automatically, or an imported chapter list.
• Lossless when the format stays the same: the pieces are the original's own encoded frames written into new files. Nothing is decoded and nothing is re-encoded.
• Track numbers and titles: every piece is tagged with its position and the total, and a chapter list gives each one its own name.
• Filename patterns: put the number, the title or the timestamp into the names so a player reads them back in the right order.
• Pick and choose: untick any piece you do not want before exporting, and the numbering still refers to the recording rather than the folder.
• One archive: save any piece on its own, or collect them all into a single download.

Splitting is one operation with five very different questions in front of it. Which one you want depends entirely on what the recording is, so the tool offers all five rather than making you work backwards from the only one it has.

• Equal parts. Say how many pieces you want and the recording divides evenly. This is the right answer for a two-sided tape, a set of exercises, or anything you want in halves or thirds.
• Fixed length. Cut every so many seconds, with the last piece keeping whatever is left over. This is what a transcription service, an upload limit or a podcast host usually asks for.
• Marks. Place each cut yourself by dragging on the waveform or typing an exact time. Nothing beats this when the boundaries are musical rather than arithmetic.
• Silence. Let the tool find the gaps. This is how an album rip, a cassette transfer, a set of voice memos or a DJ mix actually gets split.
• Chapters. Paste the boundaries you already have — a cue sheet from a rip, or the chapter list from a podcast description — and each entry becomes both a cut and a title.

Audio in an MP3, an M4A or an OGG is stored as a long run of small encoded frames. Cutting between two of them needs no understanding of the sound at all — the frames on one side go into one file and the frames on the other side go into the next. Nothing is decoded, nothing is compressed again, and the pieces contain exactly the bytes the original did.

That is what happens here whenever the output format is left as Same as source and nothing else you have asked for would change the samples. It is also why it is quick: a ten-minute recording becomes twenty pieces in well under a second, because the work is copying rather than encoding. Almost every other splitter decodes and re-encodes this case, which costs time and a generation of quality for nothing.

When the pieces are a byte copy

• The output format is Same as source.
• The bitrate, sample rate, channels and bit depth are all left alone.
• No fade is applied.
• The recording is within what your plan exports.

What a copy cannot do

A cut has to land on a frame boundary, because a frame is the smallest thing that can be copied. In practice that means a piece can begin up to a few hundredths of a second away from the exact time you asked for — inaudible, and the price of not re-encoding. If you need a cut placed to the sample, choose WAV or FLAC as the output and the pieces will be encoded to the exact times instead. The tool says which of the two it is doing, before you press anything.

A recording of an album side, a stack of voice memos or a live set has its boundaries built in: the quiet bits between them. Finding those by hand means listening to the whole thing with a finger on the pause key. The scan does it in a fraction of the running time and then hands you the marks to adjust.

The three numbers that matter

• Quieter than. How quiet a passage has to be before it counts as a gap. Around −35 dB suits most music; a room recording with audible hiss needs a higher number, a clean studio master a lower one.
• Gap at least. How long that quiet has to last. Short values find the pauses inside a sentence; longer ones find only the breaks between whole pieces.
• Pieces at least. The shortest track you are willing to end up with. This is what stops a run of short gaps producing a run of two-second fragments.

Each cut goes in the middle of the gap it found, so neither the track before nor the track after loses its tail or its run-in. If the scan finds nothing, that is a real answer rather than a failure — raise the threshold, or shorten the minimum gap, and scan again. Every mark it produces can be dragged afterwards.

The boundaries of an album, an audiobook or a podcast episode are usually written down somewhere already: in the cue sheet that came with the rip, or in the chapter list in the show notes. Retyping twenty timestamps into a form is not a thing anyone finishes, so paste the list instead and each entry becomes both a cut and a name.

What you can paste

• A real cue sheet. The TRACK, TITLE and INDEX lines a ripper writes. Times are read in CD frames — seventy-five to the second — which is what a cue sheet actually stores.
• A chapter list. One entry a line, with the time anywhere on it: 04:12 Opening, [00:04:12] Opening, or Opening — 4:12 all read correctly.
• Plain seconds. 252.5 Opening, if that is the shape your notes are in.

An entry at 00:00 names the first piece rather than creating a cut before it, so a list that starts at zero gives you the number of pieces you expected rather than an empty one at the front. Entries past the end of the recording are left out and counted, so a list that belongs to a longer version of the file tells you so instead of quietly producing fewer pieces.

Twenty files called output_3.mp3 are twenty files nobody can sort, tag or import. The filename pattern decides what the pieces are actually called, and the numbers in it are padded to the width of the total so an alphabetical listing puts piece 10 after piece 9 rather than between 1 and 2.

What you can put in a pattern

• {name} the original file's name, without its extension.
• {n} the piece's number, padded — 01 to 20, or 001 to 500.
• {title} the piece's title, when a chapter list gave it one.
• {start} and {end} the piece's position in the recording, as 04m12s.
• {total} how many pieces there are altogether.

Separately from the filename, every piece is tagged with its track number and the total, and with its title where one exists, so a music player orders them correctly even if the files are later renamed. The recording's own tags are carried across unless you ask for them to be removed. Cover art is not carried: the pieces are audio only.

Leaving the format on Same as source is both the fastest option and the only lossless one. Choosing a format instead re-encodes every piece, which is worth doing when the pieces are going somewhere the original format cannot go.

What each one is for

• Same as source. A byte copy. Nothing is decoded, nothing is re-encoded, and the pieces are identical to the parts of the original they came from.
• MP3. Plays on everything. The right choice when the pieces are going to a device or a service that will not take anything else.
• M4A and AAC. The same audio in two different wrappers — M4A for players and phones, raw AAC for tools that expect a bare stream.
• OGG. An open format, useful where MP3 licensing is a concern.
• FLAC. Compressed with no loss at all, at the source's own bit depth. The right output when the pieces are an archive rather than a listen.
• WAV. Uncompressed, and the format that needs no encoder at all. Big files, and cuts placed to the exact sample.

• An album rip into its tracks. One long file plus its cue sheet, or a silence scan where no cue sheet survived.
• A lecture or a sermon into chunks. Fixed length, so each piece fits an upload limit or a listening session.
• An audiobook into chapters. The chapter list from the description, pasted in, with each piece taking its own title.
• An interview for transcription. Five-minute pieces with a few seconds of overlap, so no word is cut in half at a boundary.
• A DJ set or a live recording. Silence detection with a long minimum gap, so only the real breaks count.
• Voice memos recorded back to back. Silence detection with a short minimum gap, which separates them without any listening.

The free plan splits one file at a time, up to 50 MB and thirty minutes, into as many as twenty pieces, exporting at up to 44.1 kHz and 192 kbps. Every mode is available on the free plan, including silence detection and chapter import — they are the reason to use this tool, and a tool nobody can try is not a tool.

Pro removes the imposed caps rather than raising them: up to twenty-five files at once, five hundred pieces each, 48 kHz, 320 kbps and 24-bit output, and no file-size or length limit beyond what your own device can manage. A recording above the free plan's export ceiling is still split on the free plan — it is encoded at the ceiling and says so, rather than being refused.

The recording is read, split and written entirely on your own machine. No part of it is sent over the network at any point, there is nothing to delete afterwards, and the tool works the same with the network disconnected. That is also why a same-format split finishes almost immediately: there is no upload to wait through and nothing to fetch back.

Powered by optimised audio processing algorithms.

• Split MP3 Files
• Split WAV Files
• Split OGG Files
• Split M4A Files
• Split FLAC Files
• Split AAC Files

• Audio cutter
• Audio merger
• Silence remover
• Audio compressor
• Ringtone maker
• Audio converter