How to Normalise Audio Volume for MP3 and M4A Files

By the M4AtoMP3.org editorial team · October 21, 2026 · 10 min read

A voice memo recorded across a room that you can barely hear. A playlist where every third song makes you reach for the volume knob. An interview where the guest is half as loud as the host. All of these are volume problems, and the usual fix is normalisation: adjusting a file’s overall level to a standard target. Done well, it’s quick and invisible. Done badly, it can make noise louder or cause distortion. This guide explains the two main types, the targets to use, and how to apply them in common tools.

Two kinds of normalisation

Peak normalisation

The tool finds the loudest single sample in the file and raises (or lowers) the whole file so that peak hits your target — typically −1 dBFS (one decibel below digital maximum).

• Good for: making a single quiet recording as loud as it can go without clipping.
• Limitation: peaks don’t reflect how loud something sounds. A recording with one sudden clap and otherwise quiet speech will barely get louder, because the clap is already near the maximum.

Loudness normalisation (LUFS)

The tool measures perceived loudness across the whole file, in LUFS (Loudness Units relative to Full Scale), using a standard (ITU-R BS.1770 / EBU R128) that weights frequencies the way human hearing does. It then adjusts the file to a target such as −16 LUFS.

• Good for: making many files sound equally loud — playlists, podcast episodes, lectures, audiobook chapters.
• Detail: a true-peak limit (e.g. −1 dBTP) is set at the same time so the result doesn’t clip.

Recommended targets

You don’t need to hit these exactly. The point is consistency: similar content at a similar level.

Method 1: Audacity

Peak normalisation:

1. Open the file and press Ctrl/Cmd+A to select everything.
2. Choose Effect › Volume and Compression › Normalize.
3. Tick Remove DC offset and Normalize peak amplitude to −1.0 dB.
4. Apply, then export as MP3.

Loudness normalisation:

1. Select all.
2. Choose Effect › Volume and Compression › Loudness Normalization.
3. Set perceived loudness to −16 LUFS (or −19 for mono speech). Tick Treat mono as dual-mono if appropriate.
4. Apply. If the waveform now touches the top and bottom, add Effect › Volume and Compression › Limiter with a limit of −1 dB.
5. Export.

For many files, put these steps into an Audacity Macro and run it on a folder — see our Audacity guide.

Method 2: ffmpeg

ffmpeg’s loudnorm filter implements EBU R128 loudness normalisation. The quick, single-pass version:

ffmpeg -i in.m4a -af loudnorm=I=-16:TP=-1.5:LRA=11 -c:a libmp3lame -b:a 128k out.mp3

• I=-16 — integrated loudness target in LUFS.
• TP=-1.5 — true-peak ceiling.
• LRA=11 — loudness range target; lower values compress dynamics more.

For the most accurate result, use two passes. The first measures:

ffmpeg -i in.m4a -af loudnorm=I=-16:TP=-1.5:LRA=11:print_format=json -f null -

It prints values such as input_i, input_tp, input_lra and input_thresh. Feed them into the second pass:

ffmpeg -i in.m4a -af loudnorm=I=-16:TP=-1.5:LRA=11:measured_I=-27.2:measured_TP=-4.1:measured_LRA=6.3:measured_thresh=-37.6:linear=true -c:a libmp3lame -b:a 128k out.mp3

With linear=true, ffmpeg applies a simple gain change where possible, preserving the original dynamics. Single-pass mode adjusts dynamically and can audibly “pump” on music.

Peak normalisation in ffmpeg takes a measurement too: run ffmpeg -i in.m4a -af volumedetect -f null - to see max_volume, then apply the difference, e.g. -af volume=5.3dB if the max was −6.3 dB and your target is −1 dB.

Method 3: MP3Gain and ReplayGain (no re-encoding)

If your files are already MP3, you can even out a library without decoding and re-encoding — so without any quality loss.

• MP3Gain (Windows; AACGain and command-line ports elsewhere) changes the global gain field in each MP3 frame in steps of 1.5 dB. The change is lossless and reversible. Use Track Gain for shuffled playlists and Album Gain to keep the relative levels within albums.
• ReplayGain tags store a volume adjustment in the file’s tags without changing the audio; compatible players (foobar2000, many Android players, Kodi) apply it on playback. foobar2000 can scan and tag a whole library quickly.

Not every device honours ReplayGain tags — most car stereos ignore them — whereas MP3Gain’s change works everywhere.

Fixing the things normalisation can’t

Normalisation changes the overall level only. Common problems it won’t solve on its own:

• One speaker much quieter than another. Use compression (Audacity’s Compressor, or ffmpeg’s acompressor) to reduce the difference between loud and quiet parts, or adjust each speaker’s section separately. Recording each person on a separate track is the real fix.
• Background hiss gets louder. Raising a quiet file raises its noise too. Apply noise reduction before normalising. In Audacity: Effect › Noise Removal and Repair › Noise Reduction. In ffmpeg: afftdn (e.g. -af afftdn=nf=-25,loudnorm=I=-16).
• Distortion that was recorded in. If the original clipped (flat-topped waveforms), normalisation can’t repair it. Clip-repair tools can help a little.
• Volume changing over time — e.g. someone walking away from the phone. Use a leveller such as ffmpeg’s dynaudnorm filter, which adapts gain over time: -af dynaudnorm=f=250:g=15.

A practical workflow for recordings

For voice memos, meetings or interviews you’re about to convert to MP3:

1. Trim silence and unwanted sections.
2. Reduce noise gently if there’s a hum or hiss.
3. Compress lightly if speakers vary a lot.
4. Loudness-normalise to −16 LUFS (stereo) or −19 LUFS (mono), true peak −1.5 dBTP.
5. Export as MP3: 64–96 kbps mono for speech, 128 kbps for podcasts with music.

With ffmpeg, steps 2–5 can be one command:

ffmpeg -i memo.m4a -ac 1 -af "afftdn=nf=-25,acompressor=threshold=-20dB:ratio=3,loudnorm=I=-19:TP=-1.5" -c:a libmp3lame -b:a 64k memo.mp3

If you only need a straight conversion with no level changes, our M4A to MP3 converter does that in the browser.

Normalising a batch of files consistently

For a folder of lectures, meetings or podcast episodes, apply the same target to each file so they all play at a similar level. On macOS or Linux:

mkdir -p normalised
for f in *.m4a; do ffmpeg -n -loglevel error -i "$f" -af loudnorm=I=-16:TP=-1.5:LRA=11 -c:a libmp3lame -b:a 128k "normalised/${f%.m4a}.mp3"; done

On Windows PowerShell, use the Get-ChildItem | ForEach-Object pattern from our batch conversion guide with the same -af option. For music albums, prefer album-level adjustment (MP3Gain’s Album Gain or ReplayGain album mode) so quiet intros stay quieter than the loud tracks around them, as the artist intended.

Checking the result

Measure a finished file with ffmpeg -i out.mp3 -af ebur128 -f null - — the summary at the end shows integrated loudness (I) and true peak. In Audacity, Analyze › Contrast and the meter toolbar give a quick visual check. Aim to be within about 1 LU of your target.

Summary

Use peak normalisation to make one quiet file as loud as it can safely go, and loudness normalisation in LUFS to make many files sound equally loud: about −16 LUFS for stereo podcasts and voice, −19 for mono, −14 for music, always with a −1 dBTP ceiling. Audacity and ffmpeg both do it in a few clicks or one command, and MP3Gain or ReplayGain evens out an existing MP3 library without re-encoding.

← Back to all articles