Make Podcast and Meeting Audio Consistent with Loudness Normalization

Step-by-step

1. 1

   Understand why peak normalization doesn't fix quiet audio

   Peak normalization finds the single loudest sample in your file and scales everything so that sample hits a target, typically 0 dBFS or just below. It says nothing about how loud the recording feels. A meeting recording where someone coughs near the microphone once already has a peak near maximum, so peak normalization changes almost nothing — the speech underneath stays quiet. That's why 'normalized' files can still sound wildly different from each other: their peaks match, but their average perceived loudness doesn't. Loudness normalization solves the actual problem by measuring how loud the material sounds over its whole duration and adjusting gain to hit a loudness target instead of a peak target.

2. 2

   Learn what LUFS means in plain language

   LUFS stands for Loudness Units relative to Full Scale. It comes from the EBU R128 broadcast standard and is designed to track perceived loudness: the measurement weights frequencies roughly the way human hearing does and averages over time, so a value of, say, −16 LUFS describes how loud the program material feels, not where its peaks land. The number most tools report and target is integrated loudness — one figure for the entire file. Values are negative because 0 is the digital maximum; a file at −14 LUFS is louder than one at −20 LUFS. One loudness unit (LU) corresponds to one decibel of gain change, which makes the math easy: to move a −22 LUFS recording to −16 LUFS, apply +6 dB.

3. 3

   Pick a sensible loudness target

   There is no single mandatory number, but there are widely used conventions. Music streaming platforms generally normalize playback to somewhere around −14 to −16 LUFS, so material delivered near that range plays back without heavy platform adjustment. For spoken-word podcasts, a common convention is around −16 LUFS for stereo and around −19 LUFS for mono — speech benefits from a slightly quieter target with more consistent dynamics. For internal meeting recordings, anything consistent in the −16 to −19 LUFS zone works; consistency across files matters far more than the exact figure. Treat these as conventions, not rules: pick one target and use it for every episode.

4. 4

   Leave true peak headroom at −1 dBTP

   Digital audio can clip in a sneaky way: the reconstructed analog waveform between samples can swing higher than any individual sample. That's what a true peak (dBTP) measurement catches. If you normalize loudness and let true peaks reach 0, a later MP3 or AAC encode can push those inter-sample peaks over the top and add audible distortion. The standard defensive setting is a true peak ceiling of −1 dBTP — quiet enough that lossy encoding won't clip, loud enough that you sacrifice almost nothing. Good loudness normalizers apply this ceiling automatically as a limiter after the gain adjustment, so a loud recording being turned up doesn't slam into 0.

5. 5

   Extract the audio track first if you're starting from video

   Meeting platforms usually hand you an MP4 video even when you only care about the audio. Extract the audio first with the MP4 → MP3 converter — drop in the recording and download the audio-only file. Working on the extracted audio is faster, and it's the file you'll actually publish or archive. If you plan to normalize, don't pick an aggressively low MP3 bitrate at this stage; keep quality high so the normalization step has clean material to work with.

6. 6

   Normalize with the Audio Normalizer

   Open the Audio Normalizer, upload the file, choose your loudness target, and process. The tool measures the file's loudness and applies the gain needed to reach the target. Afterwards, listen to three spots: the quietest speaker, the loudest moment, and a transition between them. The quiet parts should now sit at a comfortable level without you touching the volume knob. If soft passages are still hard to hear relative to loud ones, the file has a wide dynamic range problem — normalization moves the average, it doesn't compress the swings — and you may want compression at the recording or editing stage next time.

7. 7

   Trim dead air when it earns its keep

   Long silences do two bad things: they waste the listener's time, and leading/trailing silence can slightly skew loudness measurement on short files because the gating in the measurement has less speech to work with. Use the Silence Trimmer when a meeting starts with two minutes of 'can everyone hear me?' or a recording tail runs long after everyone left. For a well-edited podcast episode, you usually don't need it. When you do use both tools, trim first, normalize second, so the measurement reflects the final content.