Audio Splitter
Search for an audio splitter and page one is sites that want the file uploaded before they will make a single cut, or that give you one cut point at a time. This one decodes the recording in this browser tab with the Web Audio API, so the moment a file opens you are looking at its waveform, and the parts appear on it as you set the cuts. There are three ways to set them: the times you type, like 3:00, 7:00, a number of equal parts, or a fixed interval that keeps the short final piece rather than dropping it. Every time you type becomes a sample index once, by rounding to the nearest sample, so a cut at 3 seconds of a 44.1 kHz recording is sample 132,300 and the part before it ends on exactly that sample. Each part starts where the previous one ended, so the pieces rejoin the original sample for sample, and each is written as uncompressed 16-bit PCM WAV: download one on its own, or take the whole set as a ZIP. A split makes at most 100 parts, and the ceilings are 50 MB of file and 30 minutes of decoded audio, checked before the file is read and again before anything is cut, because the whole recording is held as samples in this tab rather than on somebody's server.
Open an audio file
· or drop one anywhere on the stage
Ready. Choose an audio file, or drop one on the stage.
Times takes the points to cut at, so 3:00, 7:00 on a ten minute file gives three parts. Equal parts divides the file into as many pieces as you ask for, and Every cuts at a fixed length and keeps the short final piece rather than dropping it. Every part is a button: click one to select it, then download that part on its own, or take the whole set as a ZIP. A split makes at most 100 parts, and the file ceilings are 50 MB and 30 minutes, because the whole recording is decoded to 32-bit samples in the memory of this tab before anything is cut.
Common questions
- Is my audio uploaded anywhere?
- No. The file is read by the decoder built into this browser, the waveform is drawn on a canvas here, and every WAV is written by JavaScript in this tab and handed to the browser's download. There is no upload, no queue, no account and no request of any kind, so no part of the recording reaches a server, an analytics call or an error report. The only things kept between visits are the settings, which is the mode, the cut times, the part count and the interval, stored in this browser under one key. The Start over button at the top of the page forgets them.
- Is there silence detection here?
- No, and that is worth being plain about, because it is the one thing people most often expect from a splitter. There is no silence detection in this tool. It does not find the gaps for you, it does not remove the quiet stretches between takes, and it never moves a cut point you set to somewhere it thinks is better. What it does instead is show you the waveform with every part marked on it, so you can see where the sound drops away and type that time yourself, then read back the exact sample each cut landed on. If a recording needs the pauses found automatically, that is a job for a desktop editor with a threshold control.
- How exact are the cut points?
- They land on the nearest sample to the time you type, and you can read the number back. A time becomes a sample index once, by rounding rather than truncating, so 3 seconds of a 44.1 kHz recording is sample 132,300 and 7 seconds is sample 308,700, and the part between them holds exactly 176,400 samples. The parts are contiguous by arithmetic and not by luck: each one starts on the sample the previous one ended on and the last one ends on the final sample of the file, so nothing is lost between parts and nothing is duplicated. Play the parts back to back and you have the original recording, sample for sample.
- Why are the parts WAV files, and why are they bigger than what I started with?
- Because browsers decode many formats and encode almost none. There is no MP3 or AAC encoder built into the platform, so a splitter that runs entirely in the page either ships a large encoder of its own or writes the one format the platform can write, which is uncompressed 16-bit PCM in a RIFF container. This tool writes the WAV. That means a part is about 10 MB a minute for stereo at 44.1 kHz, larger than the compressed file you opened, and the size of each part is printed on its card before you download anything. What you get in exchange is that the audio inside a part is never put through a second lossy encode. Two honest details sit alongside that: whatever the original compression already removed is gone before this tool ever sees the file, and the WAV is written at 16 bits, so the browser's 32-bit float samples are rounded to the nearest 16-bit value and anything past full scale is clamped rather than allowed to wrap.
- Which audio formats can I open here?
- Whatever your browser can already decode, which is the honest answer because the decoding is the browser's own. In practice that means MP3, WAV, M4A and AAC, OGG and Opus, and FLAC in current versions of Chrome, Firefox, Safari and Edge, with the exact list differing a little between them. There is no separate list of supported formats maintained here and nothing to keep up to date: the file is handed to the browser's decoder, and if it comes back undecodable the tool says so and names the file rather than failing silently. A proprietary or DRM-protected file has to be exported as a WAV or an MP3 first.
- Can I download all the parts at once?
- Yes. All as ZIP cuts every part, writes each one as a WAV and puts the set in a single ZIP file, stored rather than deflated, because PCM audio barely compresses and paying several seconds a part for it is the wrong trade in a browser tab. The parts are written one at a time with the progress shown, so a long file does not freeze the page, and Stop ends the export and leaves your file open and your cut points as they were. Each part is named after the source file with a zero-padded number, so ten parts sort as 01 to 10 in a file manager instead of 1, 10, 2. Any single part can also be downloaded on its own: select its card and use Download part.
- How many parts can one split make?
- Up to 100. It is not a paywall, since nothing is uploaded and there is nothing to charge for: it is the point past which a hundred separate WAVs and the ZIP holding them stop fitting comfortably in one tab's memory. Asking for more, whether through the part count, too many cut times or an interval that is too short, is refused with the number of parts your settings would have made and the cap named, rather than a frozen page. For more pieces than that, split the file in two passes.
- What are the size and length limits, and why does a local tool have any?
- 50 MB of file and 30 minutes of decoded audio. Neither is a commercial gate. They exist because decoding expands audio: 30 minutes at 44.1 kHz is 79.4 million frames, and in stereo that is over 600 MB of 32-bit floats held in this tab before a single byte of WAV is written, and every part written out allocates its own copy on top of that. The size is checked against the file before it is read at all, and the length is checked against the decoded audio before anything is sliced, so a file past either ceiling gets a message naming the limit and its own actual value instead of a dead tab. For anything longer, cut it into halves in a desktop editor first.
- What happens if I type a cut time that does not work?
- It is refused with the value named, rather than quietly changed. A cut at or past the end of the file says how long the file actually is, the same time typed twice says which time is repeated, a cut at 0:00 says that the first part would be empty, and anything that is not a time at all says which forms are accepted: 3:00, 1:02:30, or 45 for forty-five seconds. Cut times out of order are the one exception, because they have an obvious meaning: 7:00, 3:00 is sorted into 3:00, 7:00 and used. Nothing is cut until the whole list makes sense, and the message changes as the mistake changes.
Cuts land on the nearest sample to the time you enter, and every piece is written as uncompressed PCM WAV, so nothing is re-compressed and the segments join back to the original sample for sample. It does not find the gaps for you: there is no silence detection here, so the cut points are the ones you set.