How do you split one MP3 into several tracks without losing anything?
Short answer: in the audio cutter, add one region per track with + Add region, then type each region's start as exactly the previous region's end, set both fades to 0 and press Download next to each region. We split a 30-second recording into three tracks this way and checked the result:
- Saved as WAV: the three tracks were 441,000 samples each and added back up to the 1,323,000-sample source. No sample was lost or doubled at a split point; every sample matched the source within 1 step of 16-bit rounding.
- Saved as MP3 (128 kbps): each 10.000 s track decoded to 10.031 s, with the music starting 25.1 ms late. Played one after another, every track change gets 31.0 ms of extra silence (49.0 ms with our 15-minute tracks).
- A 90-minute MP3 split into six 15-minute tracks in Chromium on our server: the file opened in 11.8 s, the browser peaked at 2.4 GB of memory, each WAV track saved in about 2 s and each 128 kbps MP3 track in 43 to 53 s. The six WAV tracks added back up to the source to the sample.
Pick WAV if the tracks will be played back to back or joined again; pick MP3 for separate clips you will share one by one. Watch the default: + Add region starts the new region 0.5 s after the last one, so if you do not retype the start, that half second is in no track.
How splitting works in a browser
An MP3 splitter in a browser does not chop the MP3 file itself. It decodes the whole file to samples, copies the samples between each start and end point into a new file, and writes that file as WAV or as a new MP3. On this site that is the audio cutter's region list: each region becomes one track, and each one downloads as its own file.
That gives two things to check. First, whether two regions that touch (one ends at 10.000 s, the next starts at 10.000 s) share the boundary cleanly, without a missing or repeated sample. Second, what the MP3 encoder adds to each track. The page turns a typed time into a sample number by rounding time × 44,100, and both regions round the same 10.000 to the same sample, 441,000, so the first track stops just before the sample where the second begins. We measured whether the files agree.
How we tested
- Tool: this site's audio cutter, run in headless Chromium through Playwright on a local copy of the page. The script opened each file, pressed + Add region, typed the start and end times, chose the export format and clicked Download on every region.
- Music: 30 seconds of Mozart's Symphony No. 40, first movement, played by the Musopen Symphony, in the public domain on Wikimedia Commons, converted to 44.1 kHz, 16-bit stereo. We split it as a WAV and as a 128 kbps MP3 made with FFmpeg's LAME encoder, into 0-10, 10-20 and 20-30 s.
- Long recording: the same excerpt looped to 90 minutes, mono, saved as a 128 kbps MP3 (86.4 MB), the size of one side of a long tape or a lecture. Split into six 15-minute tracks.
- Checks: we decoded every downloaded track with FFmpeg 4.4.2, joined the samples end to end and compared them with the source. For MP3 tracks we found where the music starts by cross-correlation with the source. Memory is the total resident memory of all Chromium processes, sampled every 0.2 s.
Do the tracks add back up to the original?
| Source | Tracks saved as | Each track | Tracks joined vs source |
|---|---|---|---|
| WAV, 30 s | WAV | 441,000 samples (10.000 s) | 1,323,000 of 1,323,000 samples; max difference 1 step |
| MP3 128 kbps, 30 s | WAV | 441,000 samples (10.000 s) | 1,323,000 of 1,323,000 samples; max difference 2 steps from FFmpeg's decode of the same MP3 |
| WAV, 30 s | MP3 128 kbps | 442,368 samples (10.031 s), 160,496 bytes | music starts 1,105 samples (25.1 ms) late in every track; 1,368 extra samples (31.0 ms) per track |
| MP3 128 kbps, 30 s | MP3 128 kbps | 442,368 samples (10.031 s), 160,496 bytes | same as above |
| MP3 128 kbps, 90 min, mono | WAV | 39,690,000 samples (15:00.000), 79,380,044 bytes | 238,140,000 of 238,140,000 samples; max difference 2 steps from FFmpeg's decode |
WAV tracks are lossless splits of what the browser decoded. The 1-step differences come from turning the browser's decoded samples back into 16-bit numbers, and for an MP3 source the 2 steps are the difference between Chromium's and FFmpeg's decoders, not the split. Nothing went missing at the seams.
MP3 tracks are not. Each one is compressed again (see does cutting an MP3 lose quality), and each one gets the encoder's start delay plus padding to fill its last 1,152-sample frame. The tracks this site writes have no LAME tag telling players to remove that padding, so it stays in the audio. Splitting the MP3 source gave exactly the same track sizes and timings as splitting the WAV: what you get depends on the export, not the input.
Two settings that change the split points
The default region after + Add region
On our 30 s file, the first region opened at 7.5-22.5 s (the middle half of the file). The first press of + Add region added 23-28 s and the second added 28.5-30 s. A new region starts 0.5 s after the end of the last region and is 5 s long, or shorter when the file ends first. Left like that, 22.5-23 s and 28-28.5 s are in no track. Type each start so it equals the previous end; the page keeps times to the millisecond, so 10 and 10.000 are the same point.
Fades apply to every track
The fade in and fade out sliders are shared by all regions. With both at 0.5 s, our three WAV tracks were unchanged except near the ends, but at each split point both neighbouring tracks faded to silence: the joined result dropped to zero at the seam and stayed more than 3 dB below the unfaded split for 700 ms around each seam. For a track list that should play as one recording, leave both fades at 0. Use them only when each track is meant to stand alone.
Splitting a 90-minute recording
A long recording is where splitting is most useful, and where a browser tool can run out of room, because the page holds the whole decoded file in memory. Our 90-minute mono MP3 (86.4 MB) opened and split without trouble on a 4-core server (Intel Xeon E5-2699 v4) in headless Chromium:
| Step | Time | Result |
|---|---|---|
| Open and decode 90 min | 11.8 s | 238,140,000 samples, mono |
| Download 6 tracks as WAV | 1.6 to 2.1 s each | 79,380,044 bytes each, 476 MB in all |
| Download 6 tracks as MP3 128 kbps | 43 to 53 s each, 4 min 56 s in all | 14,400,783 bytes each, 86.4 MB in all |
| Memory, all Chromium processes | 252 MB with the browser idle; 2,434 MB at the peak while decoding | |
Each 15-minute MP3 track decoded to 2,160 samples (49.0 ms) more than 15 minutes; the padding depends on where the track length falls within the last frame, so it varies with the length. A stereo file holds twice as many samples in memory as a mono one of the same length, so a 90-minute stereo recording needs more than we measured here. If a phone struggles, split the recording on a computer, or export WAV first and work on shorter pieces.
Step by step
- Open the file in the audio cutter. Note the split points you want, in seconds or minutes:seconds.
- Region 1 is already there: type its start (often 0) and its end.
- Press + Add region. Type the end first, then the start, and make the start the same number as the previous region's end. Repeat for each track; the last end is the file's length.
- Check the region list: each line shows start → end and the length, so gaps or overlaps are easy to spot.
- Leave fade in and fade out at 0. Choose WAV for gapless tracks or MP3 for small files, and press Download on each region.
The page does not find silences or split at fixed intervals for you; every split point is one you set. The file names carry the start and end times, such as moz30_0-10-000_to_0-20-000.wav. Minutes are not padded with a zero, so a list sorted by name can put a track starting at 5:00 after one starting at 15:00; rename them if the order matters.
Questions
Can I split an MP3 without re-encoding it?
Not on this site: the page decodes the file, so MP3 export encodes each track again. WAV export avoids a second round of compression, at about 10.6 MB per stereo minute. Tools that copy MP3 frames instead can only split on frame edges, every 26.1 ms at 44.1 kHz.
Why is there a short silence between my MP3 tracks?
Each MP3 track starts 25.1 ms late and ends with padding, 31.0 ms extra per 10 s track in our test. Why there are gaps between joined MP3 files measures it in detail and shows how to remove it.
Can I put the tracks back together?
Yes, with the audio joiner. WAV tracks join back without a gap; untagged MP3 tracks keep the 31 ms at each seam.