Convert Audio

Whatever your browser can decode: WAV, MP3, Ogg Vorbis and Opus, FLAC, and AAC in M4A or MP4 in current Chrome, Edge, Firefox and Safari; AIFF in Safari. A file the browser cannot decode is refused with a message. The output is WAV, which the page writes itself, or MP3, which LAME encodes on your device.

Yes: choose MP3 under Format, with a bitrate of 64, 128, 192 or 320 kbps. The MP3 is encoded on your device by LAME (lame.sourceforge.net) in its JavaScript port lamejs 1.2.1, under the LGPL-3.0, which this site serves unchanged with its licences and loads only when you make an MP3; a long file can be stopped while it encodes. MP3 holds 8 to 48 kHz and one or two channels: another rate is converted first with the page's own filter (96 kHz becomes 48 kHz), and so is stereo at 64 kbps from 32 kHz up, which is written at 24 kHz, the rate LAME uses for it; a file with more than two channels is mixed down to stereo by its speaker layout (How are more than two channels made stereo?); the result line says each. At 16 to 24 kHz the bitrate is at most 160 kbps and at 8 to 12 kHz at most 64 kbps. Each MP3 starts with 1,105 samples of silence that the encoder adds (25 ms at 44.1 kHz, 23 ms at 48 kHz, up to 138 ms at 8 kHz) and its end is padded to a whole frame; nothing in the file marks them, so players play them too. Your browser decodes each MP3 back before it is offered, and one it cannot decode is not offered. WAV keeps every sample exactly and is about ten times larger: about 10 MB per minute of 16-bit stereo at 44.1 kHz, where an MP3 at 128 kbps is about 1 MB.

Choosing a different sample rate recomputes the samples at the new rate through a low-pass filter made for it (a Kaiser-windowed sinc, flat up to 90 % of half the new rate and 120 dB down from half the new rate on): everything the new rate cannot hold is removed first, so a high sound never folds back as a false lower tone (aliasing), and going up adds no mirror images. Frequencies above half the new rate are lost, which is expected when going down to 8 or 16 kHz for speech. The filter is the page's own, not the browser's resampler, so the result is the same in every browser. The file's own rate is read from its header (WAV, AIFF, MP3, AAC, M4A, Ogg, WebM or FLAC); if a header does not state it, or states an AAC core rate that HE-AAC doubles, the page names the rate it used. Keeping the rate, the channels and the bit depth (the defaults) with normalisation off copies the samples as they were read: a 16- or 24-bit PCM WAV, which the page reads itself, comes out sample for sample the same; other formats are decoded once by the browser, which decides their exact values. Mono mixes all channels by averaging them; stereo from mono duplicates the channel, and stereo from more channels mixes them by their speaker layout (the next answer).

They are mixed down the way ITU-R BS.775 mixes surround sound to stereo: left and right stay as they are, the centre goes into both sides at −3 dB, each surround, back or side channel into its own side at −3 dB, and the LFE (low-frequency effects) channel is left out. The page reads 3 channels as left, right, centre; 4 as left, right, surround left, surround right (quadraphonic); 5 as 5.0 (left, right, centre, surround left, surround right); 6 as 5.1 (left, right, centre, LFE, surround left, surround right); and 8 as 7.1 (left, right, centre, LFE, back left, back right, side left, side right): the usual speaker order of WAV files for that many channels, and for other formats the order the browser decodes them to. A WAV whose header declares another speaker order, and any other number of channels, has no layout the page can be sure of, so the page does not mix them. It says so, and offers Mono, which averages every channel, or a WAV with all of the file's channels. Where the mix would pass full scale, the whole file is lowered just enough to stay within it. The result line names the layout and any lowering. Choosing Stereo mixes this way, and so does MP3, which holds two channels at most.

It measures the loudest sample and scales the whole file so that peak sits at −1 dBFS, the same gain for every sample, so nothing clips and the balance is unchanged. It is peak normalisation, not loudness (LUFS) matching, so two files can still sound different in loudness. A silent file is left alone.

By default each file keeps the bit depth its header states: a 24-bit WAV or FLAC stays 24-bit, a 16-bit one stays 16-bit, a 32-bit file is written at 24 bits (the result line says so), and a lossy file such as MP3, which states none, is written at 16-bit. 16-bit is what CDs and most tools expect and is enough for finished audio; 24-bit keeps more headroom for editing. Choosing 16-bit for a 24-bit file rounds every sample to 16 bits, and the result line says so. A lossy source such as MP3 gains nothing from 24-bit beyond headroom, since the decoded signal has already lost detail. An MP3 has no bit depth: its bitrate sets its size and detail.

Files up to 50 MB each. The WAV that comes out is uncompressed: about 10 MB per minute of 16-bit stereo at 44.1 kHz, half as much again at 24-bit; an MP3 at 192 kbps takes about 1.4 MB per minute. Decoded audio takes 4 bytes per sample per channel in memory, and each file stays decoded while it is in the list: a 50 MB MP3 of about an hour at 44.1 kHz takes about 1.3 GB in stereo, or 0.6 GB in mono. Converting needs room for the WAV as well, and for a new copy of the samples when the rate or the channels change or the peak is normalised; on a phone that can fail. Keeping the sample rate takes a moment; changing it filters every sample, which takes longer: a few seconds for a song, a minute or more for a recording of an hour, and the page shows how far it has got.

Once the page has loaded, yes, for WAV: nothing is fetched while you convert. MP3 needs its encoder, about 530 KB, which the page fetches from this site the first time you make an MP3; offline before then, the page says it could not load the encoder.

No. The page loads a small script into your browser and every operation runs on your device. You can check this yourself in the browser's network panel: after the page has loaded there are no upload requests. Closing the tab discards the file.

No. These basic tools are free and store nothing on our servers. The only limits are the tool's own: files up to 50 MB each. Within them, the practical limit is your device's memory.

Nothing you open or type. While the tab is open, the tool pages keep a short list of the steps taken in it, in the tab's session storage: the step (page opened, started, succeeded, first success, used again or failed), the tool's name, the time, how many files were used and, for a failure, an error code; never a file name, the contents of a file or your text. It holds at most the latest 200 steps, and the browser deletes it when the tab closes. The page never sends it. Where a Report a problem button is shown, you can attach the last 20 steps to your report; nothing is attached unless you tick that box. In local storage the page keeps only choices you made: a Light or Dark theme until you choose System again, the language you picked with the EN / 中文 switch if it differs from your browser's language, and this tool in your recently used list if you switched that list on at the tools page.