What is LUFS?
What is LUFS? And what should LUFS Voiceover Levels be? Well, LUFS (Loudness Units Referenced to Full Scale) is the standard for measuring Loudness. Loudness is the perceived level of heard sound. It is not the same as peak volume level, which is the maximum amplitude of a sound wave. 0 dBFS (Decibels Relative to Full Scale) is the highest you can go up to in peak volume level in digital recording before the signal clips or distorts. LUFS, on the other hand, measures how loud something in a digital recording seems to sound. With the goal being to sound full, loud, and most importantly, good, with a large dynamic range.
That sounds like great way to measure a voiceover, right? Loudness, and LUFS, depends on the frequency, duration, and dynamic range of the sound. It’s not about driving the meters to the maximum as much as possible. However, that used to be the goal before LUFS, back in the bad old days not so long ago, of The Loudness Wars. So, let’s briefly look at how we got from then to now. Then we’ll look at how to apply LUFS Voiceover Levels in real word applications for voice over recordings.
Before LUFS: The Loudness Wars
Before LUFS, we had The Loudness Wars. They started in the mid 1990s when engineers tried to make CDs louder. Waves put out the first Brickwall Limiter with look-ahead, the L1, in 1994. Things were so over-compressed and raised to the highest level that CDs from bands like The Red Hot Chili Peppers had clipping in them. The Loudness Wars ended in the 2010s with streaming services. iTunes lead the way, demanding lower set levels to encourage better dynamic range and quality control. By 2014, Daft Punk won five Grammy Awards for their album Random Access Memories. That CD was deliberately mixed with less compression and lower levels, thereby giving it greater dynamic range.
The Loudness Wars launched without hard standards from the engineering and entertainment industries. At the advent of digital audio, there was an -18 dBFS standard, calibrated to correspond to 0 VU analog audio. There would be plenty of headroom, and you would have no danger of going over 0 dBFS in the digital realm. So far, so good! But then Sony set it to -12 on many products. Soon people were setting it to –10 and as high as -6. “Hey, as long as it’s not going over 0 dBFS, we’re good!” people said. I had many debates with video cameramen over levels in my career as an audio engineer. Everyone wanted levels as hot as possible, even if we were just recording a sit-down interview. LUFS Voiceover Levels would have helped settle the issue.
But there was no longer the same amount of dynamic range, and more and more over-compressed and over-limited mixes. This fueled The Loudness Wars. Everything was a big muddy mush through the mid and low range. And that’s still true today on today’s US radio, even though in 2017 AES set a recommended -20 to -16 LUFS for radio.
Then the biggest change away from the Loudness Wars came in 2010. The EBU (European Broadcasting Union) put out R 128, which set levels to -23 LUFS. They included a tolerance of +/- 1 LU (loudness unit) for live programs. True Peak levels should not exceed -1 dBTP (dB True Peak). Version 4.0 of EBU R 128 can be found here (r128.pdf (ebu.ch).
LUFS Standards Today
What are the LUFS standards today? The EBU still recommends -23 LUFS, as does the Audiobook Production Standard proposed by the Audio Publishers Association. Audible requires -23 to -18 LUFS. Spotify, YouTube, and Amazon use the -14 LUFS standard. Apple uses a -16 LUFS standard for music and podcasts. Note that all of these services require a maximum True Peak setting of -1 or -2 dBFS. Here is the current 2023 AES Loudness Normalization page. AES set levels to -18 LUFS for voiceover and dialog content and -14 LUFS for music.
How do I Use LUFS Voiceover Levels?
How does one set LUFS Voiceover Levels? I use REAPER, and LUFS are monitored both in the recording meters and in the rendering window when you export/save a file. In Audacity, there are two default settings for Normalization: “Perceived Loudness” of -23 LUFS, and RMS default setting of -20 dB. However, you can enter the levels as you like, such as industry accepted levels for Podcasts -16 LUFS for stereo files and -19 LUFS for mono files. In Audition, there are great tools for monitoring LUFS and matching the loudness in various files.
What does this mean in the real world compared to how one used to master voice overs? Well, when I take an uncompressed voice over with peaks around -12 dBFS to -10 dBFS, it will translate to around -26 or -27 LUFS as read in REAPER. Then, when I Normalize the file during the Render process in REAPER, I select “Normalize” and then enter -23 LUFS as the setting. What does that file then become? It then has peaks of -6 to -5 dBFS. How about if I compress the audio at a 4:1 ratio with 6dB of makeup gain and then Normalize at -23 LUFS? Then the file peaks at -8 to -6 dBFS.
Now, what happens if I Normalize a compressed voiceover file to the -14 LUFS like with Spotify and some other streaming services? That’s what you might want to do for voice over demos, or for auditions to appear louder. Then you have to be careful, because the levels could be super hot, possibly about +2 to +3 dB peaks with clipping! Instead of -14 LUFS, try -19 or -18 LUFS maximum. Or, start by not over-compressing before you set the mastering LUFS level. This isn’t like the old Loudness Wars days where you set a limiter to -3, -1, or -.1 dBFS and then crank the drive of the limiter so that the waveform looks like a solid block.
More Dynamic Range with LUFS Voiceover Levels
These extreme levels are what I’m still hearing from some of the top voice over demo producers and talents: -11 to -10 LUFS with peaks soaring above -1 dBFS. It’s time to sound the alarm, as it were, about still making voice overs with the Loudness Wars mindset. It’s outdated, but moreover, it doesn’t sound good. What makes great recordings is dynamic range and clarity.
When you mix voice over demos to be so loud, you’re losing the dynamic range and clarity. The nuance and humanity of the voice over is lost. The lows are booming, and the highs are strident. If we are going to worry about AI voice overs, we shouldn’t then be producing voice over content that has the same lack of dynamic range and nuance found in AI voices. Think about why we have the newer LUFS standards when setting LUFS Voiceover Levels. They are here to preserve the dynamic range of audio while having industry standards for levels.