MP3 Cutter
Drag a selection over the waveform, snap the edges to a zero crossing, and save the trim without the file ever leaving your device.
Drop an audio file here
MP3, WAV, M4A, OGG or FLAC. Decoded in this page, never uploaded.
Milliseconds. 3ms applies regardless.
Longer fades are an effect, not protection.
Moves each boundary up to 10ms so the cut does not click.
- Source
- —
- Selected
- —
- Result
- —
- Approx. size
- —
Why the edges are handled for you
A cut across a moving waveform is a step, and a step is an audible click. Each boundary snaps to the nearest zero crossing, and a 3 ms fade applies as a floor even with both fade fields at zero. In remove mode the fades land on the join, not on the discarded part.
Everything here runs on your device. Nothing you enter is uploaded or stored.
Most audio trimming is one small job. The first eight seconds are a chair scraping. The interview has a two-minute preamble nobody needs. A ringtone has to be thirty seconds and start on the chorus. None of that is worth a desktop editor, an account, or an upload — and an upload is the part that should bother you, because audio is often a recording of a room with people in it.
How to use it
Drop in a file. The waveform appears with the whole thing selected, along with its duration, sample rate and channel count. Drag anywhere on the waveform to make a new selection, or drag the handles at either edge, or type timecodes into the start and end fields if you know exactly what you want. With a handle focused, arrow keys move it by 10ms and Shift-arrow by a second, which matters because 10ms is far smaller than a fingertip.
Then decide whether the selection is what you are keeping or what you are throwing away, and press play to hear it before you commit. Both previews work on the real processed audio, fades included, so what you hear is what saves.
Keep and remove are different jobs
Keep is the common one: a clip, a ringtone, a quotation. The export is the selection with fades on its two new ends.
Remove is quieter and more useful. It deletes the selection and joins the halves either side of it — a cough, a phone ringing, a long pause, gone. The fades are applied to the joined result rather than to the source, so the ramp lands exactly on the seam where the two halves meet, which is the only place it does any good.
The edges are handled for you
Snap-to-zero and the 3ms fade floor are both on by default and both exist for the same reason. A cut across a moving waveform is a step, and a step is a click. Snapping moves each boundary to the nearest sample where the audio is already crossing silence, searching within 10ms so the selection you made stays the selection you meant. The fade floor covers whatever the snap could not find. Set both fade fields to zero and you still get 3ms of ramp, because a cut that ticks is a broken cut.
Longer fades are there when you want them as an effect rather than as protection — half a second in and out turns a hard clip into something that can sit under a video without announcing itself.
What comes out
WAV is written directly and instantly, with no quality change of any kind. MP3 pulls in a small encoder the first time you ask for it — around 150KB, loaded on demand rather than on page load — and offers bitrates from 128 to 320kbps. Because MP3 output means re-encoding, the page says so when your source was already lossy.
| You want | Format |
|---|---|
| The trim as a step before more editing | WAV |
| A ringtone, a voice note, something to send | MP3 at 192kbps |
| The smallest file that still sounds fine | MP3 at 128kbps |
Nothing is uploaded
The file is decoded with the browser’s own Web Audio API, trimmed as raw samples in the page, and encoded back on your machine. There is no server involved at any point, so no recording of yours is sitting in someone’s temporary folder waiting to be deleted on a schedule you cannot see. Close the tab and it is genuinely gone.
Once trimmed, Audio Converter changes the format, Audio Merger joins clips with a crossfade, and Audio Waveform Generator turns the shape you have been staring at into an image you can publish.
Questions
Why does a cut sometimes click, and how does this avoid it?
Because audio is a moving voltage, and cutting it at an arbitrary point creates a step. If the waveform happens to be at -0.4 when your selection starts, the file begins by jumping from silence to -0.4 in a single sample, and that jump is a click. Two things prevent it here and both are on by default. First, each boundary snaps to the nearest zero crossing within 10ms — a sample where the waveform is already passing through silence, so there is no step to hear. Second, a 3ms fade is applied as a floor rather than a default, so even with the fade fields set to zero the edges ramp instead of jumping. Snapping searches a mono sum of every channel rather than just the left, because a boundary that is at zero on the left and at -0.6 on the right fixes half the click and sounds like it worked.
Does cutting an MP3 reduce its quality?
Exporting back to MP3 does, and there is no way around it in a browser. The file has to be decoded to raw samples before it can be trimmed, and re-encoding those samples is a second generation of lossy compression on top of the first. It is usually inaudible at 192kbps or above, but it is real, and the page tells you when it is about to happen rather than staying quiet. Exporting to WAV avoids it entirely — the samples are written out exactly as they were decoded, at the cost of a much larger file. If the trim is a step in a longer process, WAV is the right intermediate.
What does the remove mode actually do at the join?
It deletes the selected region and butts the two remaining halves together, then applies the fades to that result rather than to the source. That ordering is the whole trick. Fading the source first would put a fade at the boundary of a region you are about to throw away, leaving the actual join untouched and audible. Fading the result puts the ramp exactly where the two halves meet. Snapping helps here too, since both sides of the join land on a zero crossing and the level is already matched.
Why are the silent gaps offered as buttons?
Because they are almost always where you meant to cut. Any run quieter than about -42dB for longer than 0.4 seconds is detected and listed as a timecode chip. Clicking one selects it, which in remove mode strips a pause out of a voice recording in two clicks, and in keep mode gives you a boundary that was already silent. It is a shortcut, not a decision — the selection stays fully editable afterwards.
How large a file can it handle?
Large, but the ceiling is your device's memory rather than any limit set here. Audio is decoded to 32-bit floating point samples, which is roughly 10MB per minute of stereo at 44.1kHz regardless of how compressed the source was. An hour-long podcast is around 600MB in memory, which a desktop handles and a phone may not. If a long file fails to load, the error says so plainly. Trimming in shorter passes is the practical answer.
Last updated