Loudness normalization conforms audio to a perceived loudness level. Learn why that's important in podcasting and how to make your podcast meet the standard.
Why a loudness normalization standard matters
Imagine you press play on a podcast episode. The intro music is a bit louder than the previous podcast you listened to, so you turn down the volume. Then, the main host comes in to introduce the topic, but they're quieter than the music, so you have to turn up the volume. Then, the cohost comes on and they're even quieter than the host, so you have to turn up the volume again or else not be able to hear them.
After a while, they segue into their interview with some bumper music, which is much louder than their voices, so you rush to turn down the volume before you damage your ears. This is a separately recorded interview and you are, again, having to continuously adjust the volume level so you can hear both participants and not damage your hearing.
At last, the podcast is over, but you have to turn down the volume one last time because the outro music is too loud. Then, the next podcast on your player starts playing, and it's too quiet, so you have to turn up the volume again, and the fight continues.
Does that sound familiar? That kind of frustration happens every day and could happen to any podcast—even the professionally produced ones!
That's why podcasts need loudness normalization! This would ensure the only time a listener must adjust their volume is when their environment changes, not when the podcast's audio changes.
This volume-fighting annoyance could be easily solved with loudness normalization in three places:
- Within each episode—ensuring the participants and sound clips are all at the same loudness level.
- Across episodes—ensuring that all episodes of one podcast are the same loudness as each other.
- Across podcasts—ensuring that podcasts from separate creators are all the same loudness.
Loudness normalization solves that by conforming all pieces of audio to the same standard.
How loudness is measured
Perceived loudness is now commonly indicated by “loudness units relative to full scale,” or “LUFS” (pronounced “luhfs”) for short. In the past, it was also called “loudness, K-weighted, relative to full scale” (LKFS), and there used to be some technical differences between LKFS and LUFS. But today, they're essentially the same—so much that whenever you see “LKFS” you can assume it also means “LUFS.”
LUFS are an absolute measurement relative to the full scale of 0 dB. Thus, you'll see LUFS indicated with negative numbers: -16 LUFS, -19 LUFS, -23 LUFS, and such. (You may also hear people abbreviate “negative 19” to “neg 19” in speech).
LUFS are an indication of the unit of measurement, which is actually “loudness units” (LU). LUs are equal units to decibels (dB). So if you need to amplify by 2 loudness units, you would simply amplify by 2 decibels.
The algorithm behind LUFS is designed to measure long-term averages of audio, not mere peaks (like most normalization tools). For example, audio with a loud but momentary peak will barely affect the long-term measured loudness of the processed section. But a standard normalizer would raise or lower the audio so that that peak reaches a target level, regardless of the rest of the audio.
Loudness normalization is, therefore, the measurement of only the average, long-term perceived loudness of audio. It actually has nothing to do with dynamic range or peaks.
However, the user experience and perception of loudness is made of three parts:
- Loudness range (LRA): the statistical difference between loud and quiet over time, measured in loudness units (LU)
- Integrated/program loudness: the perceived loudness of selected audio, measured in LUFS
- True peak: the microsampled level of the loudest point in the audio (more accurate than a normal peak), measured in dB or dBTP (decibels true peak)
Standard loudness-measurement tools will show you those three parts because they help you ensure a fully consistent loudness experience.
For example, audio with a program loudness of -16 LUFS but a loudness range of 20 LU could have so much variation that a listener still has to fight with the volume controls. Thus, the loudness range should be reduced to make it more listenable, and then the whole renormalized to the target loudness.
And although true peak is often measured in loudness tools, it actually doesn't affect our perception of the overall loudness. For example, audio at -16 LUFS with a true peak of -4 dB would sound the same as the audio with a true peak of -2 dB. That's because the peaks are so small that they are usually unnoticed. It's when audio peaks for a longer time that it becomes noticeable. Nonetheless, limiting the true peak reduces the chance for distortion.
The loudness standard for podcasts
Broadcast radio and television generally have a loudness standard of -23 LUFS (gated so it excludes measurement of audio below a reasonable threshold). But podcasts are Internet media and not broadcast radio or television. The technical considerations and environments are significantly different.
Many popular media programs already offer loudness normalization through optional features (such as “voice boost,” “sound check,” and other terms). So when many of the greatest minds in audio engineering and contributors in broadcast standards proposed a loudness standard for podcasts, they align nicely with pre-existing common practices. This not only provides a consistent experience for Internet-based media, it also accounts for the diverse environments people consume Internet-based media.
Thus, we have the standard of -16 LUFS for stereo and -19 LUFS for mono (more on the 3 dB difference in a moment). I've referred to this as a “proposed standard” for many years, but seeing the broad adoption and lack of significant support for any competing proposal, I'm now comfortable calling this the loudness standard for podcasts.
The reason for the 3 dB difference between stereo and mono is because of “pan law” (or sometimes called “pan rule”). This is a mixing and recording principle based on the physics of sound. Pan law is intended to ensure a consistent volume level if you were to pan audio between left, right, or center it across two channels. Essentially, the doubling of a mono signal on most devices will result in a 3 dB increase in perceived volume. You may see this in audio-editing software where converting a stereo track to mono, or a mono to stereo will actually change the waveform, but may not affect the perceptual loudness.
There are caveats and exceptions to this, and it seems the industry of apps and devices may never shift to fully compensate for pan law. Although that's the goal, in the meantime, we should account for pan law by making the adjustments ourselves based on the format of media we publish.
That's why the standard for mono is 3 dB lower than for stereo. Although it will measure lower with most tools (but some tools actually measure mono and stereo to the same perceptual loudness), most apps and devices will play the -19 LUFS mono audio at the same loudness of a -16 LUFS stereo audio. Paul Figgiani has a nice explanation in “Podcast Loudness: Mono vs. Stereo Perception.”
Here are two audio samples you can download and try yourself (right-click to save/download the WAV audio files):
So, the loudness standard for podcasts is -16 LUFS for stereo and -19 LUFS for mono.
In addition to that standard, there are some ideal targets for true peak and loudness range.
I recommend the true peak not be any higher than -1.5 dB. This isn't part of the standard and there no requirement for what the true peak should be, only that it's best that it not go above that limit in order to minimize the chance of distortion.
Loudness range (LRA) is something else to consider, but there's no standard on it, either, only a recommendation.
There are several factors that could require different loudness ranges. Music, for example, is designed with a wide loudness range, and you may want to keep that (especially if your music fades in or out). But there could also be cases in your podcast that you would want the music to not have as much dynamic range, such as when it's in the background and you don't want the strong contrast to conflict with the foreground voices.
In general, spoken word is probably best with a loudness range below 8 loudness units (LU). But consider the cause for the loudness range. For example, you may get emphatic during a portion of your content, and it might be important for the increased loudness to remain noticeable. But if you intend to stay at a consistent loudness and see a high loudness range, you may want to target an LRA below 6 LU or maybe even 4 LU. But if your audio starts sounding overly compressed and almost robotic, then your loudness range is probably too small.
Also, the loudness range could be affected by variations between participants, where one person is consistently quieter than the other. In such a case, it would be better to normalize the multiple sources (whether on separate tracks or within a single track) to the same LUFS target so there's not as much loudness difference between them. If the voices are on separate tracks, then you can easily normalize them without compression, and the loudness range will also improve.
Universal workflow for loudness normalization
Ready to dig in? Even if you have a tool designed for loudness normalization (which I'll cover below), it's important to understand the whole loudness normalization workflow so you can make appropriate adjustments when necessary.
The elements of loudness normalization and what affects them are as follows.
- Loudness range (LRA) is affected either by vocal technique consistency and adjustable with compression, or it's affected by mismatched loudness from different tracks and adjustable with independent normalization.
- Integrated/program loudness is affected by recording levels and adjustable with gain/amplification.
- True peak (dBTP) is affected by the loudest points in your audio and is adjustable (or distortion prevented) with limiting.
I made a video tutorial to demonstrate this process with exact details and how to use each tool. That's available exclusively to members of Podcasters' Society (if you're a member, click here to watch the tutorial). Here's a summary of the process (inspired by Paul Figgiani's “Podcast Loudness Processing Workflow”).
- Measure loudness range with r128x-GUI for macOS or Orban Loudness Meter for Windows.
- Compress if necessary to reduce the loudness range.
- Measure LUFS to see how offset they are.
- Adjust gain to an intermediate target of -24 for stereo, or -27 for mono.
- Hard limit to -9.5 dBTP, which is best with a true-peak limiter, but a standard limiter will probably be okay.
- Amplify by 8 dB.
- Remeasure to ensure you're at the target.
- Adjust gain if necessary, but a small variation in loudness is acceptable.
These steps, along with the free measuring tools, will work with any audio-editing software, such as Audacity, GarageBand, and more.
However, you may get better results or save more time by using tools designed for this exact thing.
Easy tools for loudness normalization
- r128x-GUI (free, macOS): for measuring loudness, LRA, and dBTP.
- Orban Loudness Meter (free, Windows): for measuring loudness, LRA, and dBTP.
- Auphonic (free and paid options; web, Windows, and macOS): my favorite tool for processing audio quickly and easily. You can set the target loudness, enable the adaptive leveler to reduce LRA, and it has an automatic true-peak limiter. Auphonic can also reduce background noise.
- Adobe Audition's included “Match Volume” (paid, Windows and macOS): loudness normalization built right into my preferred digital audio workstation (DAW). The Match Volume tool works on files or multitrack clips and can measure and adjust to a customizable target, with built-in true-peak limiting. However, it doesn't measure or affect LRA, so you may need to do that separately.
- Adobe Audition's included “Loudness Radar” (paid, Windows and macOS): not for normalizing, but for visually measuring. This will show you the loudness, LRA, and dBTP of any played audio. It even works if you use the playback shuttle to play your audio faster!
- Hindenburg Journalist and Journalist Pro: the best standalone, single-purchase DAWs for podcasting, in my opinion! Both Journalist and Journalist Pro can automatically loudness-normalize any clip you drop into the editor. Then, you can export with a target loudness preset.
- FixMyLevels (free and paid options, web): the new kid on the block in loudness normalization. FixMyLevels has more aggressive algorithms than Auphonic, and they have designed it specifically for podcasts. It's free to try during the beta stage and seems it will cost less than Auphonic once launched.
- Standing Water Studios Simple & Mighty (SWS / S&M) extensions for Reaper (plugin is free, Reaper is paid, Windows and macOS): measurement and normalization within the popular and inexpensive Reaper DAW. More information here.
There are other tools for measuring and adjusting audio, too. Either through a plugin, standalone app, or built-in feature. With these measurement tools, “online” refers to during playback, while “offline” can analyze the audio faster than real time and without playback.
Thank you for the podcast reviews!
- Zachary Webb wrote in iTunes USA, “Daniel provides great information on the Audacity to Podcast. The way he shares information is easy to understand. Thanks for doing this podcast, Daniel!”
- Sean Hiland, host of Slept in Government Class, wrote in iTunes USA, “Utility, Clarity, Sincerity: These are the three words I first think when I listen to Daniel's episodes. I always hear something I can use and never have to filter through garbage to get to the gold. Daniel also presents a clear desire to help people improve at something he truly loves. When I was first starting out, I listened to Daniel AFTER I paid for a website service my new podcast didn't need. Thanks to Daniel, I was able to get a refund before the trial period ran out. That's real utility in a format I can enjoy.”
Need personalized podcasting help?
Ask your questions or share your feedback
- Comment on the shownotes
- Leave a voicemail at (903) 231-2221
- Email feedback@TheAudacitytoPodcast.com (audio files welcome)
Connect with me
- Subscribe to The Audacity to Podcast on Apple Podcasts or on Android.
- Join the Facebook Page and watch live podcasting Q&A on Mondays at 2pm (ET)
- Subscribe on YouTube for video reviews, Q&A, and more
- Follow @theDanielJLewis
This post may contain links to products or services with which I have an affiliate relationship and may receive compensation from your actions through such links. However, I don't let that corrupt my perspective and I don't recommend only affiliates.