What do you lose when you convert stereo to mono?
Short answer: converting stereo to mono adds the left and right sides together and halves the result. Anything that is the same on both sides comes through unchanged; everything else gets quieter, and sound that is opposite on the two sides disappears. We ran test files through the Mono setting of this site's WAV to MP3 converter, which writes exactly (left + right) / 2:
- Same on both sides (a centred voice): 0.00 dB change.
- Only on one side (a mic on the left side only): 6.02 dB quieter, peak -12.00 dBFS to -18.02 dBFS. Volume at 200% brought it back to -12.00 exactly.
- Different on each side (wide stereo, room sound): 3.02 dB quieter.
- Opposite polarity (one side flipped): silence, every sample zero. A 170° phase difference left -21.2 dB.
- One side 1 ms late: notches at 501, 1502, 2506, 3507 and 4511 Hz (the first three measured 43 to 45 dB deep), a hollow sound.
- A real orchestra recording: 1.17 dB quieter overall; by octave the drop ran from 0.56 to 2.27 dB.
Settings: on WAV to MP3 set the stereo/mono menu to Mono and pick MP3 or WAV output. If the sound was on one side only, also set Volume to 200%. If the mono file loses something, one side is probably flipped; this site cannot flip one side, so use the FFmpeg command below.
Why the mix loses anything
Mono has only one signal, so the left and right sides must be combined into one. The usual way is to average them: each output sample is the left sample plus the right sample, divided by two. Halving keeps a sound that is identical on both sides at its old level and stops the sum from clipping.
A stereo signal can be split into two parts: the part both sides share (mid) and the part where they differ (side). Averaging keeps the mid and throws the side away. A sound panned fully left is half mid and half side, so half of it, 6.02 dB, goes. Two unrelated sounds on the two sides add up in power rather than in waveform, so they lose 3 dB. A sound with one side flipped upside down is all side, so nothing of it is left. A small delay on one side makes a sound partly mid and partly side depending on its frequency, so some frequencies cancel and others pass.
How we tested
- Page: a local copy of this site's WAV to MP3 page in Chromium 153 driven by Playwright 1.63, set to Mono, 44.1 kHz, output WAV 16-bit or MP3 128 kbps. The files are what the browser downloaded. For every WAV output we checked each sample against (left + right) / 2 of the source: none was more than 1 step (1 LSB) away, which is rounding.
- Test files, written by a script as 10-second 16-bit 44.1 kHz stereo WAVs: a 1 kHz tone at -12 dBFS peak on both sides, on the left side only, and in both with the right side shifted by 45 to 180 degrees; independent white noise on each side (-20 dBFS RMS); the same noise with the right side flipped; the same noise with the right side 44 samples (0.998 ms) late; and a 440 Hz tone on both sides plus noise that is flipped between them.
- Music: 30 seconds (1:00 to 1:30) of Mozart's Symphony No. 40, first movement, played by the Musopen Symphony, public domain on Wikimedia Commons, as a 16-bit 44.1 kHz WAV, and the same excerpt as a 192 kbps MP3 made with FFmpeg 4.4.2.
- Level change is the RMS level of the mono file minus the RMS level of the stereo source (both sides averaged), from 1 second in to 1 second before the end. MP3s were decoded with FFmpeg 4.4.2.
What each kind of stereo loses
| Source | What it stands for | Mono level change |
|---|---|---|
| 1 kHz tone, same on both sides | a voice or instrument in the centre | 0.00 dB |
| 1 kHz tone, left side only | a recording with one mic on one side | -6.02 dB (peak -12.00 to -18.02 dBFS) |
| Same, volume 200% | the fix for the line above | 0.00 dB (peak -12.00 dBFS) |
| Independent noise on each side | wide stereo, room sound, applause | -3.02 dB |
| Noise, right side 0.998 ms late | two mics at different distances | -3.00 dB overall, with notches (next table) |
| Noise, right side flipped | a mis-wired mic or cable | silence (all 441,000 samples zero) |
| 440 Hz tone in both + flipped noise | a centred voice over a widened pad | tone -15.01 dBFS RMS before and after; noise -20.0 dBFS RMS to gone |
| Mozart, 30 s | an ordinary stereo recording | -1.17 dB (peak -2.96 to -4.33 dBFS) |
The last-but-one row is the one to watch in music. Stereo wideners and some synth pads put part of their sound in opposite polarity on the two sides. It sounds big in headphones and is simply not there in mono, while the centred voice next to it keeps its exact level.
Phase: how far apart the two sides can be
The same 1 kHz tone on both sides, with the right side shifted by a growing amount. The formula for two equal tones, 20 × log10 |cos(shift / 2)|, predicts the result to within 0.01 dB.
| Shift | Measured | Formula |
|---|---|---|
| 0° | 0.00 dB | 0.00 dB |
| 45° | -0.69 dB | -0.69 dB |
| 90° | -3.01 dB | -3.01 dB |
| 120° | -6.02 dB | -6.02 dB |
| 150° | -11.74 dB | -11.74 dB |
| 170° | -21.20 dB | -21.19 dB |
| 180° | silence | silence |
Up to 90° the loss is small. Past 120° it falls fast, and near 180° almost nothing is left.
A small delay between the sides
When the same sound reaches one side a little later, as with two mics at different distances from a source, each frequency gets a different phase shift. A 0.998 ms delay is half a cycle at 501 Hz, so 501 Hz cancels; one cycle at 1002 Hz, so 1002 Hz passes; and so on up the spectrum. The result is a row of evenly spaced notches, which is why such a mix sounds hollow or thin.
| Frequency | Measured | Formula |
|---|---|---|
| 100 Hz | -0.4 dB | -0.4 dB |
| 250 Hz | -3.0 dB | -3.0 dB |
| 501 Hz | -44.9 dB | -56.3 dB |
| 750 Hz | -3.0 dB | -3.0 dB |
| 1000 Hz | 0.0 dB | 0.0 dB |
| 1505 Hz | -43.3 dB | -48.3 dB |
| 2000 Hz | 0.0 dB | 0.0 dB |
| 2506 Hz | -45.2 dB | -62.3 dB |
| 5000 Hz, 10000 Hz | 0.0 dB | 0.0 dB |
A longer delay moves the first notch lower and packs the notches closer together. Overall the noise came out 3.00 dB quieter, the same as unrelated sound, so a level meter will not tell you the mix sounds hollow; listening will.
Music, MP3 output and MP3 input
In the Mozart excerpt the two sides had a correlation of 0.527 and the side part was 5.09 dB below the mid part. Mono came out 1.17 dB quieter. By octave the loss ran from 0.56 dB (250 to 500 Hz) to 2.27 dB (31 to 62 Hz), so the balance of the recording barely changed. The mono file matched the mid part within 1 LSB.
- MP3 output: every mono MP3 at 128 kbps was 0.45 dB quieter than the mono WAV of the same source. A stereo MP3 at 128 kbps from the same encoder (lamejs) lost 0.45 dB on the tone and 0.41 dB on the Mozart against its stereo WAV, so that small drop comes with MP3 encoding and decoding, not with the mono mix.
- File size: a mono WAV is half the stereo WAV (30 s: 2,646,558 against 5,292,558 bytes). An MP3's size is set by its bitrate, not by whether it is mono or stereo; choosing Mono lets you pick bitrates down to 64 kbps at 44.1 kHz. How much smaller each option makes a WAV has the full table.
- MP3 input: the converter's file picker lists only WAV files, but an MP3 dropped onto the box converted to mono fine. The page labels it "Not a WAV file we can read the header of" and then converts it anyway; the dropped and picked files gave identical output.
What to do when mono loses something
- Convert once to mono WAV and listen. If the voice or instrument you care about is still there at the same level, you are done; pick MP3 if you need a small file.
- Sound only on one side: keep Mono and set Volume to 200%, which adds exactly 6.02 dB back. The page warns "Above 100%, loud parts can clip": anything on that side louder than -6 dBFS will clip, so lower the volume until it doesn't.
- Something vanishes or sounds hollow: one side is flipped or late. This site's converter only averages the two sides and cannot flip one. The free command-line tool FFmpeg can. In our test (FFmpeg 4.4.2) the flipped noise that went silent in a plain mix came back at its full -20.01 dBFS with:
ffmpeg -i in.wav -af "pan=mono|c0=0.5*c0-0.5*c1" out.wav - One side is useless (a dead or noisy side): keep only the good one. This command kept the left side bit for bit in our test:
ffmpeg -i in.wav -af "pan=mono|c0=c0" out.wav
Usec0=c1for the right side.
Questions
Does converting to mono lower the quality?
It removes the difference between the two sides and nothing else. For a centred source the WAV output matched the source to within 1 LSB. What you lose is the width, plus whatever the side part held, which in the Mozart test was 1.17 dB of level.
Why is my mono file quieter than the stereo one?
Because part of the sound was on one side or different between the sides: one-sided sound loses 6.02 dB and unrelated sound 3.02 dB. The converter divides by two so the sum cannot clip; raise the Volume slider if you need the level back.
Can I turn a mono file into stereo here?
Yes. On the WAV to MP3 page, the Stereo setting copies a mono file into both sides, so the sound stays centred. It does not add any width.