Audio Tools

What these tools are for

This category covers the small jobs that sit between recording something and actually using it. Audio Trimmer cuts a clip down to the part you want. Audio Joiner puts several clips end to end. Audio Converter changes a file from one format to another. Voice Recorder captures straight from your microphone when you just need a quick take.

The rest handle the fiddly bits. Silence Remover strips dead air out of a long recording. Volume Normaliser evens up a clip that drifts too quiet. Noise Reduction pulls down steady background hiss. Audio Speed & Pitch slows a passage or shifts its key. Musicians tend to reach for Metronome, Tuner and BPM Counter, while Tone Generator and Audio Spectrum Analyser are for testing gear and seeing what a signal really contains. Karaoke Vocal Remover has a go at lifting the lead vocal out of a finished mix.

What changes when the work happens in the tab

Editing sound here means the file is decoded into raw samples and held in memory before you touch it. That is why trimming and joining feel immediate: the whole clip is already loaded, so nothing has to be read again. The trade-off is memory. A very long recording becomes a large block of data, and a modest machine can start to struggle, so it is often easier to work in shorter pieces.

The live tools behave differently again. Metronome, Tuner and Tone Generator run on the browser's audio clock rather than the page's drawing loop, so the beat stays steady while the page is busy with something else. The microphone tools ask permission the way any site does, and the sound goes into the page and no further. One honest limit: which formats you can read and write depends on what your browser itself supports, and that varies a little between browsers.

Before you start

Order matters more than people expect. Clean first, then level. Run Noise Reduction and Silence Remover before Volume Normaliser, because normalising a hissy recording lifts the hiss along with the voice. Convert last of all: every pass through a compressed format costs a little quality, so edit once and export once.

Two more things worth knowing. Quiet and noisy are different problems, and reaching for Volume Normaliser when the real fault is background noise usually makes a recording sound worse. And be realistic about Karaoke Vocal Remover. Pulling a voice back out of a finished mix is genuinely hard, and the result normally carries some artefacts, so test it on one chorus before you build an evening around it.

Why a WAV saves at once and an OGG takes as long as the song

A browser can read far more audio than it can write. Hand it an MP3, M4A, OGG, FLAC, WAV or Opus file and it turns the whole thing into raw samples in a moment. Ask for a file back and the picture changes. The only encoder a browser gives a web page is MediaRecorder, and MediaRecorder works by recording a live stream, so writing WebM/Opus, M4A/AAC or OGG/Opus means the clip is played through from beginning to end. A four minute track takes four minutes. The export panel prints that estimate in seconds before you commit to it.

Two of the three routes avoid the wait. WAV is written here by hand, straight from the samples: a RIFF header and then the numbers, in 16-bit, 24-bit or 32-bit float. MP3 has an encoder of its own, LAME, a 156 KB script served from this site and fetched the first time you ask for an MP3, then kept in the browser cache. It is a script that runs on your machine, not a service you send audio to. It is also several times faster than the clip is long, which is why MP3 and WAV are the two sensible choices for anything longer than a few minutes.

Peak, average and LUFS are three different answers

Volume and Normalise works on the peak. It finds the single loudest sample and scales the whole file so that sample lands just under full scale, a decibel down by default. That is a promise about distortion, not about loudness. One stray click can hold the peak at maximum while everything else on the track is quiet, which is why the same page also shows the root-mean-square average, and shows both figures before and after your change.

The Loudness Meter answers the other question properly. It follows ITU-R BS.1770-4: the K-weighting filter, mean square over 400 ms blocks stepped every 100 ms, an absolute gate at -70 LUFS and then a relative gate 10 LU below the mean of what survived, with loudness range from EBU Tech 3342 on top. Given the EBU Tech 3341 compliance signal, a 1 kHz sine at -23.0 dBFS on both channels, it reads -22.99 LUFS, inside the 0.1 the test allows. It then sets your file beside the numbers people actually deliver against: -14 LUFS for most music streaming, -16 for spoken word and podcasts, -23 for EBU R128 broadcast, -24 for ATSC A/85 and -27 for cinema.

The meter also separates true peak from sample peak, and the gap between them is worth knowing. Between any two samples the reconstructed wave can rise higher than either of them, so a file measuring a comfortable -0.1 dBFS on its samples can still overload a converter or an encoder. The meter rebuilds the signal at four times the sample rate to find those overshoots and prints the difference in dB. If a file is heading for MP3 or AAC afterwards, that number is the argument for leaving some headroom rather than normalising flat to the top.

Four tools cut audio, and the difference is who chooses the cut

Audio Cutter is the one where you choose. Drag a selection across the waveform, keep it or delete it, step back through the last twenty edits if you change your mind. Every edit happens on the decoded samples, so it lands immediately however long the file is, and the export is the only slow part.

Silence Remover chooses by level. It measures the recording in 10 ms windows, short enough to fall between words, marks every run that stays under your threshold for long enough to count as a gap, then rebuilds the file with those gaps shortened or gone. Split a Recording into Tracks chooses from one of three things: the quiet passages, a tracklist or CUE sheet you already have, or plain equal parts. It returns numbered WAV files in a ZIP, with each cut nudged to the nearest point where the wave crosses zero so the edges do not click, and each track's title written into a LIST INFO chunk inside the file, so they arrive named instead of as track 1 through track 12. If you feed it a CUE sheet, the times in it are counted in frames, seventy-five to the second.

Ringtone Maker chooses by the destination. Tell it what the clip is for and it caps the length accordingly, 30 seconds for an Android or iPhone ringtone, 60 for an alarm, 5 for a notification, adds the fade, and then explains the step nobody writes down: for Android, save the M4A and copy it into the Ringtones folder; for an iPhone, save the M4A, rename it to .m4r and add it through Finder.