How to Trim an Audio File Online, and How to Read a BPM Result

AudioUpdated 2026-09-29

Two small jobs sit next to each other more often than you would think: cut a track down to a usable length, and find out how fast it is. Audio Clipper does the first, BPM Detector does the second, and both behave better when you know what they are doing under the hood.

Trimming: what the tool is actually doing

Load a file and the browser decodes it into raw samples, draws a waveform from them, and edits at that level. There is no "keyframe" problem the way there is with video, because audio has no keyframes — every cut lands exactly where you put it.

That is the useful property here. Cut points are computed at sample resolution, not rounded to the nearest frame or the nearest tenth of a second shown in the input box. If you are trimming on a beat, a boundary that sits a few milliseconds late is audible when the clip loops, and this is the kind of tool that avoids it.

The limits worth knowing:

50 MB per file. Checked before anything is decoded, with an explicit error message.
The picker filters to MP3, WAV, FLAC, OGG, M4A and AAC. A file the browser cannot decode simply will not load, so exotic containers are out regardless of what the filter says.
Output is always a 16-bit PCM WAV. There is no format selector. Whatever you feed in — MP3, M4A, a 24-bit FLAC — what comes out is uncompressed 16-bit.
Channel count is preserved. Stereo in, stereo out.

The information panel reports a bit depth of 32 for the loaded file. That describes the decoded samples sitting in memory, which are floating-point; the WAV written to disk is 16-bit. Do not expect a 32-bit export.

Setting cut points

There are two ways in, kept in sync: Start and End sliders, each paired with a numeric field in seconds at 0.1 steps. Typing a value moves the slider; dragging it rewrites the field. The waveform above is drawn on a canvas with a shaded overlay marking the selected span, and the readout beneath shows the range to the millisecond. Values outside the file's length are pulled back to the nearest legal end rather than rejected.

Loading a new file resets the range: start goes to 0 and end goes to the full duration of the new track. So a saved "0:12–0:47" from the previous file is not silently applied to the next one — but the range is also not remembered when you want it to be, so note your numbers down if you are doing the same cut across several files.

The exported file is named with the range in it, in the form song_clip_12.5s-47.0s.wav, which makes a folder of takes legible without opening anything.

What trimming does not include

This is a cutter, not an editor. There is no fade in or out, no volume envelope, no silence detection, no normalisation, and no crossfade between two regions. The edges of a cut are hard.

That matters in one practical case: a hard cut at a non-zero amplitude point clicks. The fix is to nudge the boundary a few milliseconds toward a quieter part of the wave, and the tool supports that loop directly — the Preview button plays only the selected range, not the whole file. Move an edge, preview, repeat until the clip starts and ends where you want it. For fades and normalisation you need a real audio editor.

If you need the trimmed clip smaller than the WAV it produces, convert it afterwards — see converting audio in the browser for what the OGG path costs in time.

Detecting tempo: why the number is often "wrong"

The BPM detector runs autocorrelation, and understanding one sentence about it explains almost every surprising result: an autocorrelation cannot tell a beat from its own multiples. A pulse that repeats every 0.5 seconds is genuinely periodic at 0.5 s and at 1.0 s. Both readings are mathematically correct; one of them is what musicians call the tempo.

What the tool does with a file:

Decodes it in the browser, takes the left channel only, and limits the analysis window to the first 120 seconds.
Downsamples to roughly 11 kHz — plenty for rhythm, since tempo lives in energy fluctuations, not in pitch content.
Builds a loudness envelope from short overlapping windows, then compares that envelope against itself at every offset. Offsets that align well are candidate periods.
Normalises candidates into a 55–210 BPM working range and returns the top five, each with a confidence score.

Consequences of those choices:

Half-time and double-time answers are normal. If you expected 90 and got 180, look at the candidate list — 90 is usually there. This is the single most common "error" and it is not an error.
Confidence is relative, not a correctness score. It measures how strongly that periodicity stood out against the other candidates. A metronomically rigid track can score very high on the half-time reading of itself.
Only the first two minutes are seen. A song that changes tempo in the third verse is described by its opening. A DJ mixing two keys can be misled by a long silent or ambient intro.
Only the left channel is seen. Irrelevant for almost any finished recording; a real problem for a mix where the percussion is panned hard right.

The check that actually works: audition

Numbers are the weak part of this tool; the metronome is the strong part. The audition button plays sixteen clicks at the detected tempo, with a higher tone on every fourth beat, so the bar line is audible. Clicking any candidate in the list auditions that candidate instead.

Play it against the track. Within four or five clicks you will know whether the downbeat lands with the music. That is a better test than any percentage, and it is the step people skip.

Practical notes

Files need at least about a second of audio before there is enough data to analyse; a fragment that short returns noise.
The detector accepts the same 50 MB ceiling as the clipper, and the analysis itself takes a couple of seconds at most — it is the decode that costs time on a long file.
Nothing is uploaded. Both pages decode the file in the tab and only your finished result reaches your disk, though these pages carry the site's usual analytics and advertising scripts.

Convert audio files in your browser — what WAV and OGG output really cost you in size and time
Capturing video frames and making a GIF — the same local-decoding idea applied to picture
Batch resize and convert images — when the job is many files, not one long one