When you need WAV instead of MP3
WAV is uncompressed audio: every sample stored as a number, nothing to decode. That makes it the format that works everywhere. CD burning software writes audio CDs from it, DJ software and hardware samplers load it without a decoding step, video editors and digital audio workstations prefer it, and some upload forms for radio, podcasts and course platforms ask for it. It is also the right format to edit in, because saving an edited MP3 as MP3 again compresses it a second time.
What WAV can't do is make an MP3 sound better. The MP3 encoder threw part of the sound away to make the file small, and decoding can't bring it back. The WAV is a perfect copy of what the MP3 sounds like, and much larger:
| One minute of stereo audio | Size |
|---|---|
| MP3, 128 kbps | 0.96 MB |
| MP3, 320 kbps | 2.4 MB |
| WAV, 16-bit, 44.1 kHz | 10.6 MB |
| WAV, 24-bit, 48 kHz | 17.3 MB |
| WAV, 32-bit float, 48 kHz | 23 MB |
Which settings to choose
| The WAV is for | Bit depth | Sample rate |
|---|---|---|
| Burning an audio CD | 16-bit | 44.1 kHz |
| DJ software, samplers, older hardware | 16-bit | 44.1 kHz |
| A video project | 24-bit | 48 kHz |
| Editing or mixing in an audio editor | 24-bit or 32-bit float | Same as the MP3 |
| An upload form that just says “WAV” | 16-bit | Same as the MP3 |
16-bit is CD quality and is already more than the MP3 holds. 24-bit gives an editor headroom to change the volume without adding noise, at half as much again in size. Changing the sample rate is only worth it when the software or the CD needs a particular one; most MP3s are 44.1 kHz already.
The silence at the start of every MP3
An MP3 encoder can't start encoding on the very first sample, so it adds a short stretch of silence at the start and pads the end to fill the last frame. With the LAME encoder that is 1,105 samples at the start, 25 milliseconds at 44.1 kHz. Players hide it by reading a gapless header that LAME writes into the first frame, which records exactly how much was added.
We checked what the conversion does with it, using a 20-second test recording of exactly 882,000 samples and two MP3s made from it:
| MP3 | Length of the WAV | Start compared with the original |
|---|---|---|
| With a gapless header | 882,000 samples, exact | same sample |
| Without one | 883,584 samples | 1,105 samples (25 ms) late |
With the header, the WAV lines up with the original sample for sample, which is what you want for loops, DJ cue points and audio laid against video. Without it there is no reliable way to tell the encoder's silence from the start of the music, so it is left alone, and each result says which kind of MP3 it came from.
MP3s that already clip
Loud, heavily limited music often comes out of MP3 decoding slightly louder than it went in, and the highest peaks go past full scale. The browser's decoder cuts them off, as most players do. We made a 320 kbps MP3 from a master limited to exactly full scale: decoded, 1,876 samples in 20 seconds sat at the limit, while a decoder that doesn't clip measured the peaks at 3.5 dB over.
A WAV can only hold what the decoder gives it, so a 24-bit or float WAV doesn't bring those peaks back. The tool counts them and tells you when the number is large enough to matter, so a crackle on a loud passage isn't blamed on the conversion. If you are making MP3s yourself, WAV to MP3 and FLAC to MP3 lower the level just enough to stop it happening.
Tags, albums and privacy
Title, artist, album, year, genre, track number and comment are copied from the MP3's ID3 tag into the WAV's own tag list, which Windows and many audio editors read. Cover art has no standard place in a WAV and is left out. Drop a whole album and download it as one ZIP named after it.
The MP3 is decoded and the WAV is written in your browser; nothing is uploaded. Long files need memory: an hour of audio is about 635 MB as a 16-bit WAV, which a laptop handles and an older phone may not. To shorten a file first, use Trim audio.