Convert to MP3
Turn anything into an MP3. Pick your quality and go.
or paste one from your clipboard
- Length
- ·
- Sample rate
- ·
- Channels
- ·
- Size
- ·
- Length
- ·
- Sample rate
- ·
- Channels
- ·
- Size
- ·
If your source is already an MP3, matching its bitrate keeps the size similar without paying a second quality penalty.
| Bitrate | Per minute | What it is for |
|---|---|---|
| 64 kbps | 0.46 MB | Speech only. Music sounds hollow and swirly. |
| 96 kbps | 0.69 MB | Voice memos, lectures, audiobooks. |
| 128 kbps | 0.92 MB | The old CD-ripping default. Audible on cymbals. |
| 160 kbps | 1.1 MB | Comfortable for casual music listening. |
| 192 kbps | 1.4 MB | The sensible default for music. Hard to fault. |
| 256 kbps | 1.8 MB | Effectively transparent on most material. |
| 320 kbps | 2.3 MB | The MP3 ceiling. Use when storage is irrelevant. |
MP3 is constant bitrate here, so the size is fixed by the bitrate and the duration alone. A quiet acoustic track and a dense mix at 192 kbps come out identical in size.
How to use it
- Drop a WAV, M4A, OGG, FLAC, Opus or WebM file onto the page. It is decoded in the tab, not uploaded anywhere.
- Read the bitrate table below and decide how much quality you are willing to trade for size.
- Set the bitrate in the Quality box in the export bar. The estimated MP3 size updates as soon as you change it.
- Click Download. The LAME encoder runs in your browser and writes the MP3 directly to your downloads folder.
Why MP3 refuses to die
MPEG-1 Audio Layer III was standardised in 1993, out of work done at Fraunhofer IIS and the University of Erlangen through the late 1980s. By any technical measure it has been beaten repeatedly since: AAC does the same job in about two thirds the bits, Opus is better again below 96 kbps, and both handle transients and high frequencies more gracefully. None of that has dented MP3's position, because compatibility compounds. A format that has been supported by every player, phone, car stereo, DJ controller, smart speaker and set-top box for a quarter of a century is a format you can send to anyone without asking what they own.
The patents that once made MP3 encoders a licensed product expired in 2017. That removed the last practical argument against it for software authors, and it is why an MP3 encoder small enough to load in a web page exists at all.
What the encoder is throwing away
MP3 compression is built on auditory masking. When two sounds occur close together in frequency and one is much louder, the quieter one becomes inaudible. Your ear genuinely does not register it. The encoder splits the signal into 576-sample blocks, transforms each into a frequency representation, estimates how much noise each band can hide, and allocates bits accordingly. Parts of the signal you could not have heard are stored coarsely or not at all.
The model breaks down in two well-known places. Sharp transients, a castanet, a rimshot, a plucked harpsichord string, can smear noise backwards in time across the block that contains them, an artifact called pre-echo, audible as a faint whoosh just before the hit. And at low bitrates the encoder often applies a low-pass filter, discarding everything above roughly 16 kHz, which is what makes a 128 kbps file sound slightly dull on cymbals and sibilance compared with the original.
When MP3 is the wrong choice
Skip MP3 if you need more than two channels, a sample rate above 48 kHz, or reliable gapless playback across a whole album. Skip it for spoken word that has to be as small as possible, where Opus at 32 kbps beats MP3 at 96 kbps outright. Skip it if the file is going into a Unity or Godot project, where OGG Vorbis is the expected asset format and MP3 decoding is awkward.
Skip it, above all, as an intermediate step. If you plan to trim, normalize or add effects afterwards, keep the working copy lossless and encode to MP3 once, at the very end. Each lossy round trip compounds, and by the third or fourth it stops being subtle.