Audio Normalization: LUFS, dBFS, and Streaming Platform Standards Explained

The Problem Normalization Solves

Before streaming platforms applied loudness normalization, the music industry experienced the "loudness war" — mastering engineers competing to make their releases sound louder than competitors by using heavy dynamic range compression. The result was music that was perceptually louder but also more fatiguing to listen to, with a crushed, distorted quality on heavily compressed tracks.

Streaming platforms solved this by normalizing all content to a consistent loudness target. Every song, podcast episode, and video on a given platform is adjusted to approximately the same perceived volume before it reaches your ears. This means making your track louder than the target achieves nothing — the platform turns it back down. And it means content below the target is turned up, which amplifies noise if the noise floor is high.

Understanding the target means your audio sounds exactly as intended on every platform, without any surprise adjustments.

dBFS: What It Measures and Its Limitations

dBFS (decibels relative to full scale) measures the amplitude of individual audio samples relative to the maximum possible digital value. 0 dBFS is the maximum — any sample at or above this causes digital clipping (distortion). Audio is always at negative dBFS values; −6 dBFS means the sample is at half the maximum amplitude; −20 dBFS is significantly quieter.

Peak normalization sets the highest single sample in the file to a specified dBFS level, typically −0.1 dBFS (just under the maximum). This prevents clipping but tells you nothing about perceptual loudness.

The limitation of peak normalization: Consider a gun shot in a quiet film scene. The gunshot is a single loud sample at 0 dBFS, but most of the audio is at −30 to −40 dBFS (nearly silent). Peak-normalizing this to −0.1 dBFS achieves nothing useful — the single loud transient is already at the maximum, and everything else remains quiet. The perceived loudness of the overall audio is still very low.

Another example: heavily compressed pop music where every sample is near the peak level results in the same 0 dBFS peak as the gunshot example, but the perceptual loudness is enormously higher because the average level is much closer to the peak. Peak normalization cannot distinguish between these cases.

When to use peak normalization: For audio where you need to guarantee no clipping occurs (setting a true peak ceiling of −1 dBFS before uploading to any platform). This is a safeguard, not a primary loudness control.

LUFS: Perceptual Loudness Measurement

LUFS (Loudness Units Full Scale) measures perceived loudness over time, accounting for the fact that human hearing is not equally sensitive to all frequencies and that brief loud sounds feel less loud than sustained loud sounds at the same peak level.

LUFS implements the ITU-R BS.1770 standard, which applies frequency weighting (boosting the 1–4 kHz range where human hearing is most sensitive, de-emphasizing very low and very high frequencies) and integrates loudness over time (ignoring brief transients, measuring the sustained loudness of the content).

Key LUFS concepts:

Integrated LUFS (I-LUFS): The average loudness over the entire file, gating out sections where the audio is more than 10 LU below the loudness already measured. This is what streaming platforms measure and normalize to. When someone says "normalize to −14 LUFS," they mean integrated LUFS.

Short-term LUFS (ST-LUFS): Loudness measured over a 3-second window, useful for monitoring sections of a longer piece.

Momentary LUFS (M-LUFS): Loudness measured over a 400ms window. Useful for mixing decisions in real time.

LRA (Loudness Range): The difference between the loudest and quietest sections of the audio (in LU). A high LRA indicates high dynamic range (classical music, film audio). A low LRA indicates compressed, consistently loud audio (pop music, podcasts).

True Peak: The maximum sample value after intersample reconstruction — slightly higher than the measured sample peak due to the way digital audio is reconstructed. True peak must be limited to −1 dBFS to prevent distortion in the D/A converter and during AAC/MP3 encoding.

Platform Loudness Targets

Each streaming platform applies its own normalization algorithm. These targets are based on the ITU-R BS.1770 standard and are publicly documented.

Music platforms:

Podcast platforms:

Video platforms:

Practical implications:

• Upload louder than the target → platform turns you down
• Upload quieter than the target → platform turns you up (also turns up noise)
• Upload at the target → your audio plays exactly as intended
• If you target −14 LUFS and your audio goes to Apple Music (−16 LUFS target), Apple Music turns you down 2 LU — acceptable
• If you target −16 LUFS and your audio goes to Spotify (−14 LUFS target), Spotify turns you up 2 LU — fine for clean audio, but exposes noise

How to Normalize: Step by Step

For a podcast episode targeting −16 LUFS:

1. Complete all editing (trim, silence removal, noise reduction)

2. Open Volume Normalizer

3. Select LUFS normalization mode (not peak)

4. Set target: −16 LUFS integrated

5. Enable true peak limiting at −1 dBTP

6. Process and download

7. Verify: open a LUFS meter or use the tool's readout to confirm the output is at −16 LUFS

For a music track targeting Spotify/YouTube (−14 LUFS):

1. Complete mixing and mastering (compression, EQ, limiting already applied)

2. Normalize to −14 LUFS integrated

3. Check that true peak does not exceed −1 dBTP (add peak limiting if needed)

4. If your mix is already louder than −14 LUFS: submit as-is (platform turns it down); or reduce the master volume to hit −14 LUFS

5. If your mix is quieter than −14 LUFS: normalize up to −14 LUFS (platform no longer boosts; your audio is full-level)

The multi-platform strategy: Normalize to −14 LUFS for Spotify/YouTube distribution; create a separate −16 LUFS version for Apple Music and podcasting if your content goes to both. This is a two-export workflow, each taking seconds.