Convert WAV to MP3

A WAV stores every sample as a raw number with nothing clever applied, which is why its size depends almost entirely on its length. CD-quality stereo runs about 10 MB per minute whether the minute holds an orchestra or silence. Forty minutes is 400 MB, and an hour-long interview is around 600.

That collides with almost every limit that exists. Most email providers stop at 25 MB. Voice-note fields, submission portals, podcast hosts and messaging apps all sit well below the size of an untouched WAV. The format is not doing anything wrong — it was designed for editing, not for sending — but the file will not move, and that is what brings people here.

A phone voice memo, a Zoom recording, an interview and a lecture are speech, and speech is exceptionally cheap to encode. Balanced writes 160 kbps, at which a spoken recording is essentially indistinguishable from the original, and Small file at 96 holds up too. An hour of speech is about 72 MB at 160 and 43 MB at 96, down from 600.

Music needs more room. Balanced at 160 kbps is the setting worth taking for anything with instruments in it, and High quality at 320 is where the argument stops for all practical purposes. The one case that genuinely rewards the higher setting is live and acoustic material — audience noise, cymbals and long reverb tails are exactly what the encoder finds hardest.

This is a newer version of the same wall and it catches people unprepared. ElevenLabs, Suno and most text-to-speech exports hand back WAV because it is the format that loses nothing on the way out. A few minutes of generated narration is therefore tens of megabytes before anyone has done anything with it.

For that material the choice is easy: generated speech is speech, so the Small file band is enough and the result is what you actually publish or embed. Keep the WAV only if you intend to edit it further — once it is mixed and final, the MP3 is the useful copy and the WAV is just storage.

MP3 is lossy by design. It models what human hearing cannot detect — a quiet sound immediately after a loud one, frequencies masked by others nearby — and discards it. The removed audio is genuinely gone, and no later conversion brings it back.

What it leaves alone is the part people worry about most. Length, pitch, speed, and the intelligibility of speech are untouched. A conversation converted at 96 kbps transcribes exactly as well as the original, which matters if the recording is going to a transcription service afterwards.

Every MP3 encode discards a little audio, so encoding an MP3 again compounds it. If you need three versions at three bitrates, make all three from the WAV rather than making a small one from a large one.

The same rule decides what to archive. Keep the WAV if the recording will ever be edited, remixed or re-cut; it is the only copy that still holds everything. If it is a voice memo you will listen to once and file, converting and deleting the WAV is a perfectly reasonable thing to do with 600 MB.

A great many WAVs are stereo files carrying one microphone, duplicated across both channels. That is common in Zoom recordings, phone memos and single-source interviews. Encoding that as stereo spends twice the bitrate to say the same thing twice.

Where the source is genuinely one voice, mono at 96 kbps sounds identical to stereo at double the rate and produces a file half the size. Keep stereo for music, for anything with a stereo field worth preserving, and for recordings with two people on separate channels.

WAV has no real metadata convention. Some tools write a small INFO chunk, most write nothing, and the fields MP3 uses — title, artist, album, artwork — usually have no source to come from. Expect the MP3 to arrive unlabelled.

If the files are going into a music library or a podcast feed, plan on tagging them afterwards; any audio player or tag editor will do it. Nothing is lost in the conversion, because there was nothing there to lose.

Drop up to a hundred WAVs and each is encoded separately, then handed back together as a ZIP. This is the normal case rather than the exception — recorders, conference tools and voice-memo apps produce sessions in batches, not one file at a time.

Because nothing is uploaded, a large batch costs no bandwidth and waits in no queue. It costs processor time on your own machine, which is why a folder of long recordings is worth starting before you make coffee.

The encoder runs inside the browser tab, so the audio is never transmitted. Given what people convert here, that is not a small detail: interviews under embargo, medical and legal recordings, therapy sessions, and voice notes recorded in someone’s home.

It also means there is no upload to sit through and no daily allowance. The one ceiling is the free tier’s 100 MB a file, which at ten megabytes a minute is about ten minutes of CD-quality stereo — a 600 MB interview has to be split before it will convert.