MP3 Quieter Than the Video? Why and How to Fix

An MP3 sounds quieter than its video mostly because of what happens to the channels, not because MP3 is a quieter format. If the video has a 5.1 or other multichannel track, MP3 can only hold two channels, so the sound is folded down to stereo, and dialogue that lives in the center channel drops by 3 dB in the process. The other usual causes are a source track that was already quiet, and a player that boosts video playback but not music files. Work through them in that order: check the source level, check the channel count, then normalize the result. [Convert MP4 to MP3](/mp4-to-mp3) extracts the track with a fixed quality setting and no volume change, so any level difference you hear was already decided by the source or by the downmix, and you can correct it afterwards.

We read the extraction code to see exactly what runs. The MP4 to MP3 tool calls ffmpeg with -vn -acodec libmp3lame -q:a 2: drop the video, encode with LAME at quality 2 (variable bitrate, roughly 170 to 210 kbps). There is no -ac option, no volume filter and no loudness normalization anywhere in that command. That means the tool never makes audio quieter on purpose, and it never makes it louder either. Two things can still change the level: LAME can only write one or two channels, so ffmpeg inserts its default downmix to stereo when the source has more, and lossy re-encoding adds small errors. Everything else you hear comes from the source file or your player. Knowing the command is useful because it tells you which causes to rule out and which to fix.

We measured it. We built 10 second 5.1 test files in ffmpeg (AAC, 384 kbps) with the same 300 Hz tone placed in different channels, ran the tool's exact arguments, and compared loudness with ffmpeg's ebur128 filter. Tone in the center channel only (where film dialogue usually sits): sample peak fell from -27.6 dB to -30.6 dB, exactly 3.0 dB lower, while measured loudness stayed at -32.3 LUFS because it was now spread over two speakers. Tone in the surround channels only: loudness fell from -27.8 to -32.3 LUFS, 4.5 LU quieter. Tone in the front left and right channels only: no change at all, -29.3 LUFS before and after. So a 5.1 movie with dialogue in the center plays about 3 dB softer in each ear after conversion, which is clearly audible. This is one synthetic tone, not a real film mix, so treat the numbers as the size of the effect, not a forecast.

Downmixing does not always lower the level, and that is worth knowing before you assume it does. In our test, a tone present in all six channels at once came out louder: loudness rose from -24.7 to -21.7 LUFS, a 3.0 LU increase, and the peak went from -27.4 to -19.8 dB. That happens because the channels are summed into two. A real film mix is somewhere between our extremes: loud effects in the fronts and surrounds may sum up, while dialogue in the center is the part that gets reduced. This is why the MP3 can feel quieter on speech yet no different on music or action. If your clip is mostly talking, expect the quieter result. If you are unsure, compare the two files with a loudness meter rather than by ear, because small differences are easy to misjudge.

If the original video was already quiet, the MP3 will be just as quiet; extraction copies the level faithfully. Our stereo control test came out at -32.3 LUFS in the MP4 and -32.3 LUFS in the MP3, with a peak of -30.8 dB in both, so a plain stereo source loses nothing measurable at this quality setting. Then there is the player. Many video players, phones and streaming apps apply volume leveling or a boost to video that they do not apply to a bare MP3 played from a file manager. Comparing a video in one app with an MP3 in another is not a fair test. Play both files in the same player at the same system volume, and check whether sound enhancement or normalization is switched on in only one of them. Often that alone explains what you hear.

Measure both files so you know which cause you have. In ffmpeg, run ffmpeg -i file -af ebur128=peak=true -f null - and read the integrated loudness (I) and the peak. Run it on the video and on the MP3. If the two match, the difference is the player or your ears. If the MP3 is lower, count channels with ffprobe -show_entries stream=channels. A result of 6 means a 5.1 downmix is the cause. If the source itself sits around -30 LUFS or lower, it was quiet from the start. As a rough guide, streaming speech is commonly delivered around -16 LUFS, so a measured value far below that explains why it feels weak. Our quiet test tone measured -32.3 LUFS, which is very quiet. Checking takes a minute and saves you from boosting a file that did not need it.

Fix it in two steps. First, downmix on purpose so the center channel is not left behind. A pan filter such as pan=stereo|FL=FC+0.5*FL+0.5*BL|FR=FC+0.5*FR+0.5*BR adds the center at full level instead of 0.707. On our center-only test file this brought the MP3 from -32.3 to -29.3 LUFS and the peak back to -27.6 dB, matching the source. Second, normalize with loudnorm: -af loudnorm=I=-16:TP=-1.5:LRA=11 took the same quiet file to -16.0 LUFS with a peak of -14.2 dB. Loudnorm raises everything to a target, so it also lifts background noise in a very quiet recording. If you only need a simple lift, [Change volume](/volume-changer) can raise a file up to 300%, but it can clip loud peaks, so listen afterwards. Our converter has no setting for either step, so run these on the result.