Universal Loudness Normalization Workflow
Loudness normalization is the process of adhering your audio to proposed loudness standards. This is primarily measured in LUFS, which stands for Loudness Units relative to Full Scale (this is essentially interchangeable with LKFS, which stands for Loudness, K-weighted, relative to Full Scale). The number you’ll see in LUFS is similar to a decibel volume level, but with a larger algorithm behind it to measure perceived loudness. Other aspects of the measurement are loudness range (or LRA), which is the variation between your loud and quiet audio; and true peak (or dBTP), which is the accurate micro-sampled level of your loudest audio points.
Normalizing your audio loudness ensures your audience will hear your audio properly and they won’t have to frequently adjust their listening volume within a single episode, across multiple episodes, or across separate podcasts that follow the proposed standards.
The best and easiest tool for loudness normalization is Auphonic’s web or desktop leveler, but there are also specialized plugins and professional audio-editing software with these tools built in, such as Adobe Audition.
Overview
- Measure loudness range
- Compress
- Measure LUFS
- Adjust to intermediate target
- Hard limit to -9.5 dBTP
- Amplify by 8 dB
- Remeasure
- Adjust gain if necessary
These steps will work with any audio-editing software and any compliant loudness measuring tool on any platform. The particular tools I’ll show you are completely free.
For measuring the audio loudness, I recommend r128x-GUI for OS X or Orban Loudness Meter for Windows. These aren’t the best tools, but they are free and do the job adequately.
For processing and editing the audio, I recommend Audacity and its built-in effects. What I’ll show in Audacity will be similar to any other audio-editing software.
Step 1: Measure loudness range
Start with your raw, recorded audio. If you're not using a hardware compressor, then you will probably need to compress your audio with software in order to decrease the loudness range. This will ensure a more consistent volume level, especially if you have multiple participants.
To get a general idea of how much compression you need, look at the loudness range (or LRA) in the measurement tool. For r128x-GUI on OS X, drag the audio file onto the program. For Orban Loudness meter on Windows, switch to the analysis tab and drag your audio into the side panel. (You’ll do this same thing whenever you need to measure your files.)
The only number you need right now is the loudness range (or LRA). For spoken word, you generally want the LRA to be lower than 8. If it's higher than that, you probably need to compress the audio.
Step 2: Compress
Audio compression is an art all by itself. Here’s the basic method to use for your spoken-word tracks, but skip this for music tracks.
In your audio-editing software, play some of the audio and watch the output volume meters. Look for the low end of the voice’s volume and that number will be the threshold, minus a couple dB. Then, look at areas that are supposed to be silent, and see how loud those sections are and that will be the noise floor. Only pay attention to “room noise,” not to additional noises (such as coughs).
Select the audio and run a standard compressor. Enter the approximate quiet number from your voice as the threshold, minus a couple dB. This tells the software to compress everything above that level. Then, enter how loud the “silence” was in the noise gate, if applicable.
The ratio is usually best around 4:1, but you may need a larger number if you have a wider loudness range.
Run the compression without make-up gain and not compressing based on peaks. The will make your audio quieter, but that’s okay. Export this audio as WAV or AIFF.
Step 3: Measure LUFS
Remeasure the new audio and look at the “Program Loudness” or LUFS number. Find the offset between that number and an intermediate target of -24 for stereo, or -27 for mono. For example, if your target is -27 and your audio measured at -33.5, then the offset is +6.5 dB.
Step 4: Adjust to intermediate target
Go back to your audio-editor. Increase or decrease the audio gain by the offset amount in order to reach the appropriate intermediate target.
The reason for this intermediate target is to protect the audio peaks from clipping. Some editing apps, such as Audacity, will clip the audio beyond repair if you over-amplify it, even if you haven’t saved the audio. Running a limiter after over-amplification still results in corrupted audio. So this intermediate target gives enough margin to adjust things as necessary without the risk of corrupting your audio.
Step 5: Hard limit to -9.5 dBTP
Now, run a limiter on your audio. This might be called simply “Limiter” or “Hard Limiter.” It’s best if there’s a True Peak setting, but that’s very not common yet, unless you’re using a professional app.
Set the hard limit to -9.5 dB. And don’t apply any make-up gain or boost. If your limiter has a True Peak setting, enable that. If you can, set the look-ahead time to 12 ms and release time to 200 ms.
Step 6: Amplify by 8 dB
Still in your audio-editing software, add 8 dB of gain. Export this again as WAV or AIFF.
Step 7: Remeasure
Analyze your new audio with your loudness measurement tool. Your Integrated/Program Loudness should be at (or very close to) -16 LUFS for stereo or -19 LUFS for mono, your Loudness Range (LRA) should be lower than 8, and your True Peak (dBTP) should be at -1.0 or lower. These results don’t have to be absolutely precise. It’s acceptable for your loudness to be off-target by a few decimal points.
Step 8: Adjust gain if necessary
If you are working with multiple audio tracks, remember that layering them, especially spoken word over background music, will affect the levels of your mix down. If the result in step 7 is off by only a few decimal points, it’s probably acceptable.
But if you want to be precisely on target with your mixed-down episodes, then adjust the overall gain up or down by the measured offset of your current LUFS and the stereo or mono target you want to reach.
When you’re finished, you’ll have a file with the right loudness, loudness range, and without any clipping distortions.
I recommend doing this process for every audio track in your episode. The one exception is that I usually don’t recommend compression on music because music is supposed to have a dynamic range.
This loudness normalization workflow is much simpler with professional software, but what I just showed you will work with nearly any audio-editing software on any platform.