Cut a piece out of a recording SNIP

Why not just use an online audio cutter

Because of what the recordings are. The three minutes of a two-hour meeting that mattered. A voice memo. An interview. A note somebody left on your phone. People reach for an audio cutter at exactly the moments when the file is a record of a real conversation, and the ordinary way to trim one is to upload the whole thing to a site funded by advertising and download the part you wanted.

The upload is the entire file, not the part you keep. So the two hours you were trying to cut down to three minutes are now on somebody else's disk, in full, under a retention policy you did not read.

Here the browser decodes the recording with the Web Audio API (MP3, M4A, WAV, OGG, WebM, and FLAC in most browsers), draws the waveform from the samples, and writes the selection back out as a file. No request is made because there is no server that would have anything to do with it.

What comes out, and why it is a WAV

The export is 16-bit PCM in a WAV container, written here by hand. That is the format everything opens, on every operating system, with no codec to install and nothing to go wrong.

It is not an MP3, and the reason is worth saying plainly rather than hiding: writing an MP3 needs an encoder library, which would be a large download to ship to everybody who opens this page. The trade is that the file is bigger and it is lossless: the samples that come out are the samples that went in, with no second generation of compression artefacts on top of whatever the original already had.

For a clip you are keeping or sending on, bigger and lossless is usually the right answer. If you need an MP3 specifically, convert the WAV afterwards with whatever the destination needs, and you will only have lost quality once.

The fades and the levelling are not decoration

Cut a recording at an arbitrary point and the waveform jumps from wherever it was to zero. You hear that as a click, every time, and it is the single most recognisable mark of an amateur edit. A fade of a few milliseconds at each end removes it, costs nothing audible, and is the reason the controls are there.

Evening out the volume matters when the clip will be listened to alongside others, or on a phone speaker. A quiet passage that was fine inside a two-hour recording with context around it is inaudible as a thirty-second clip on its own.

One thing to be careful of: those same fades smooth exactly the discontinuities that show a recording has been cut. That is a feature when you are making a clip and a problem when the recording is evidence. SPLICE shows what an edit leaves behind, so you can see what somebody else would find. If a recording matters as a record, keep the original untouched and work on a copy.

Around it: HUSH bleeps or garbles named words, VOICEGUARD disguises a voice, SCRIBE transcribes on the device, and AVSCRUB strips the metadata the original was carrying.