WAV to MP3 converter

What changes when a WAV becomes an MP3

A WAV file is, almost always, a short header followed by uncompressed PCM audio: one number per sample, per channel, with nothing thrown away. Its size is set by the sample rate, bit depth and channel count, not by what was recorded: a minute of silence costs as much disk as a minute of orchestra.

CD-quality audio is 44,100 samples a second, 16 bits each, on two channels. Multiply that out and you get 1,411 kilobits per second, or a little over 10 MB for every minute. An MP3 at 192 kbps is about 1.4 MB a minute, roughly a seventh of the size. It gets there by perceptual coding: the encoder splits the sound into frequency bands, works out which parts would be masked by louder sound nearby, and spends its bits on what it expects you to hear. The rest is discarded, and nothing brings it back.

The size maths, minute by minute

• WAV, 16-bit, 44.1 kHz, stereo (CD quality): a little over 10 MB.
• WAV, 24-bit, 48 kHz, stereo (a common recorder and video setting): about 17 MB.
• WAV, 16-bit, 44.1 kHz, mono: about 5.3 MB.
• MP3 at 128 kbps: just under 1 MB.
• MP3 at 192 kbps, the default here: about 1.4 MB.
• MP3 at 256 kbps: about 1.9 MB.
• MP3 at 320 kbps: about 2.4 MB.

This converter encodes at a constant bitrate, so an MP3’s size is its bitrate multiplied by its length, whatever the WAV was: a mono voice memo and a 24-bit stereo mix of the same duration come out the same size. These are decimal megabytes; the converter’s file list counts in units of 1,024, so its sizes read about 5% lower.

Choosing a bitrate: speech is not music

For speech — interviews, lectures, meeting recordings, voice notes — choose 128 kbps and Mono. It is the lowest setting on offer and already generous for a single voice, whose intelligible content sits well below the frequencies any of these settings remove. Mono does not make the file smaller, since the bitrate fixes the size, but it changes where the bits go: encoding white noise through the same build of LAME this page uses, 128 kbps stereo was cut off at about 17 kHz and 128 kbps mono at about 20 kHz.

Keep the original channels when they carry different things. If an interview was recorded with each speaker on their own channel, Mono mixes them together for good, and an editor can no longer treat the two voices separately.

For music, 192 kbps is a sensible default for listening copies: in the same measurement it kept everything up to about 18.5 kHz, higher than most adults can still hear. Choose 256 or 320 kbps when the file is going somewhere its quality will be judged — a demo, a mix sent for approval, a set you will play through a large system. Even at 320 kbps, the highest bitrate the MP3 standard defines, the encoder discards information; if nothing may be lost, keep the WAV or use a lossless format such as FLAC.

When to keep the WAV instead

• While you are still editing. Every time an MP3 is decoded, edited and encoded again, a second encoder discards its own share on top of the first. Edit the WAV and make the MP3 last. The audio trimmer cuts a WAV and exports WAV, so trimming first and converting here costs one lossy step, not two.
• As the master. A recording cannot be made twice. Keep the WAV as the archive copy and treat MP3s as the copies you hand out; storage is cheaper than a retake.
• When someone asked for WAV. Mix engineers, broadcasters and sample libraries often specify uncompressed audio, and an MP3 converted back to WAV does not meet that brief.
• For loops and gapless albums. An MP3 comes out slightly longer than its source — by a few hundredths of a second with this encoder — because the encoder adds a short delay at the start and pads the end out to a whole frame. Players trim that only when the file carries a header saying how much there is, and this converter does not write one. On a spoken recording it makes no difference; on a loop, or a live album split into tracks, you will hear the join.

Converting a batch of recordings

Drop every file onto the converter at once, or select several in the file picker. Set the bitrate and channels once and press Convert all. Each file gets its own row with its size and duration, a progress bar while it encodes and, when it finishes, the MP3’s size and a Download button. Download all saves each finished MP3 as a separate file, though your browser may first ask whether the page may save several files. A batch keeps files separate — to combine recordings into one file, use the audio joiner.

Files are converted one at a time, deliberately. A WAV is decoded into 32-bit floating-point samples before it is encoded, so a 16-bit file needs at least twice its own size in memory, and working in sequence keeps that bounded by the largest file rather than the whole batch. Cancel stops the batch at the file in progress: finished files keep their MP3s, and Convert all carries on from where it stopped.

Each MP3 keeps its original name with .mp3 on the end, tidied so it is safe to save: spaces and characters such as brackets become hyphens, so “Interview 03 (final).wav” downloads as “Interview-03-final.mp3”. Accents are simplified and letters outside the Latin alphabet are dropped, so a name written entirely in another script downloads as “audio.mp3”. Two files that would end up with the same name get -2, -3 and so on.

Size and length limits, in minutes of WAV

Each file can be up to 200 MB, and above 50 MB the converter warns you first, because the decoded audio has to fit in the tab’s memory all at once. For WAV, where the format fixes the size, that limit translates neatly into running time: just under 20 minutes of CD-quality stereo, about 12 minutes of 24-bit 48 kHz stereo, and just under 40 minutes of 16-bit mono at 44.1 kHz.

There is also a limit on length. The decoded recording is copied to the encoder in one piece, and that copy has to stay under 2 GB, because Chrome refuses one much larger: about 85 to 90 minutes of stereo, or 170 to 185 minutes of mono, depending on whether your browser decodes at 48 or 44.1 kHz. Most WAVs reach 200 MB long before that. A file recorded at a low sample rate is the exception: it is small on disk but still decoded at your browser’s rate, so an 8 kHz telephone recording can run past three hours inside 200 MB, and the length limit is the one it meets. Choosing Mono doubles that limit for a stereo file, and the file’s row says when it would help. For a two-hour meeting recorded at CD quality, a desktop encoder such as LAME itself or ffmpeg, which works through a file in pieces, is the better tool.

What is lost, exactly

• The top of the frequency range. LAME filters out high frequencies before encoding so its bits go further. With the build this page uses, stereo files lost everything above about 17 kHz at 128 kbps, 18.5 kHz at 192, 19.5 kHz at 256 and 20 kHz at 320.
• Detail the encoder judged inaudible. Masked sound is thrown away, not stored more compactly, and at lower bitrates sharp attacks such as hand claps can pick up pre-echo, a faint smear of noise just before the transient.
• Possibly, your WAV’s sample rate. Browsers decode audio at the sample rate of your sound output, not the file’s. A 44.1 kHz WAV opened on a machine whose output runs at 48 kHz is resampled while decoding and comes out as a 48 kHz MP3. A 96 kHz WAV comes out at 48 kHz or lower, which costs nothing audible, since the MP3 format tops out at 48 kHz anyway.
• Bit depth. An MP3 has no bit depth of its own, so a 24-bit source is a cleaner starting point rather than something the MP3 preserves. The converter does not normalise or boost anything: bring a very quiet recording up in an editor before converting, not after.
• Metadata. Title and artist fields in a WAV’s INFO chunk, a Broadcast WAV’s description and timecode, and cue markers are all dropped. The converter writes no ID3 tags or cover art of its own, so tag the MP3 afterwards in your music library or a tag editor.
• Channels beyond two. An MP3 holds one or two channels, so a file with more is mixed down, as described below.

None of it comes back by converting in the other direction. An MP3 turned into a WAV is the MP3’s sound in a larger file.

WAV files with more than two channels

Surround mixes and field recorders both produce WAVs with more than two channels, and the converter chooses how to fold them down from the channel count alone:

• Four channels are treated as quad: front left, front right, then the two surrounds. Keep original mixes channels 1 and 3 to the left and 2 and 4 to the right, at half level; Mono averages all four.
• Six channels are treated as 5.1 and get the standard speaker downmix: the centre channel, usually the dialogue, goes into both sides, and channel 4, the LFE or subwoofer feed, is left out in both modes. If the mix would clip, the file is turned down just enough to avoid it, and its row says by how much.
• Any other number: Keep original keeps the first two channels, and Mono averages them all.

So the converter cannot tell a 5.1 mix from six separate microphones. If your channels are individual tracks from a field recorder, the speaker downmix is the wrong one: tracks 1 and 3 of a four-track file end up together on the left, and track 4 of a six-track file is dropped. Export the tracks you want as a stereo or mono WAV in an audio editor such as Audacity first, and convert that.

Which WAV files will open

WAV is a container, and nearly everything in it in the wild is plain PCM: 16- or 24-bit integers, sometimes 32-bit floating point. In Chrome, which we checked, all of those decode, 96 kHz files included, as do 8-bit files and the μ-law and A-law encodings used in telephony. Some WAVs hold compressed audio instead, ADPCM being the usual example, and Chrome will not decode those; the file’s row says so rather than producing a broken MP3. Other browsers draw their own lines, so if one refuses a file, try another.

The file picker suggests WAV files, but anything your browser can decode will still convert, with a short note that it is not a WAV. For iPhone voice memos there is a dedicated M4A to MP3 converter, and screen or browser recordings go through the WebM to MP3 converter.

How it works without uploading anything

Each file is read from your disk with the browser’s File API, decoded by the browser’s own audio decoder through the Web Audio API, and handed to LAME, the long-standing open-source MP3 encoder, compiled to WebAssembly and running in a Web Worker so the page stays responsive. The finished MP3 exists only in the tab’s memory until you press Download. To check rather than trust, open your browser’s developer tools on the network tab while a file converts: the requests are for this site’s own code, the MP3 encoder among it, and blob: entries, which are the tab reading your own file locally. None of them carries your audio.

Where VoiceSnap Pro fits

This converter is a free tool from the team building VoiceSnap Pro, a voice-to-text dictation app for macOS and Windows. Hold one keyboard shortcut, speak, and punctuated text appears in whatever field your cursor is in — an email, a Slack message, a document — and every dictation is saved to a searchable notes library. It will be a one-time $39 purchase, not a subscription.

The overlap with a WAV converter is the pile of recordings behind it: voice notes and meeting captures kept because typing them up was going to take an evening. If the words are what you are after, the audio to text tool transcribes a recording on your own device, WAV included. VoiceSnap Pro is meant for the next time, so the thought becomes text as you say it. It has not launched: there is no download yet, only a waitlist. Join the waitlist for a single email on release day, and in the meantime the microphone test will tell you whether the input you record with is doing its job.