In this video,
we're going to look at DaVinci Resolve's audio
normalization feature on the Deliver page, introduced in version 18.6. We'll discuss what normalization is, how it works, and why
you might want to use it. You'll find this new feature
in the Audio tab on the Deliver page. To enable audio
normalization, check "Normalize Audio" and choose the standard that
you want to use to normalize your audio. Now that we know where to find
the tool, let's talk about how it works. Audio normalization is the process of applying gain or
attenuation to an audio recording to adjust its volume
to a target or normal level. Here's a typical example. Let's say I've set
the target volume at -14 dBFS. If I had a super quiet recording, normalization would
increase its volume until the loudest sound in that recording measures -14. If I had a super loud recording, normalization would
decrease its volume until the loudest sound is -14. But it gets a
little more complicated than that because there's
different ways to measure the volume of an audio recording. A typical volume meter
measures the peak intensity of a sound. Let me give you an example. Most audio that
video editors work with is broken up into 48,000 samples per second. If just one single
sample of that audio is super loud, a peak
meter will report that level, even if every other sample in
the measured range is super quiet. Peak meters are super useful
when recording and editing audio. For example, when
recording a sound because your priority is to
avoid clipping the signal. However, peak meters are less helpful
when it comes to mastering a final delivery. For that, it's
better to use a loudness meter. Why? Loudness is perceptual. The way we
perceive the loudness of a sound is dependent on
its frequency and duration. For example, a low frequency will
seem quieter than a sound at a higher frequency that has the same
volume measured in decibels. In the example we just mentioned
of a single frame of audio that is loud, the incredibly
short duration of that intense signal means it
won't sound loud to the viewer. As loudness meters take account of
the way your mix is perceived by the audience, they are more suitable for use
when mastering a mix for final delivery. In summary, because there are two primary methods for measuring
volume, intensity, and loudness, there are two different
standards that you can normalize audio to. Normalization helps you to deliver mixes with more consistent
volume or loudness levels. Even if your video sounds good in
isolation, it might be too loud or too quiet in comparison to videos
that come before and after it. Normalization can
prevent you from delivering mixes that are either
too quiet or too loud then. Now that we have a
better understanding of normalization, let's go
back and look at those options. We'll focus on just a few. Sample peak is the
simplest form of normalization. Volume is measured
in decibels full scale. Remember though that your signal can
still clip if it's normalized to zero decibels or near zero decibel values because there might be
intrasample peaks that won't have been revealed by a peak sample meter. If you are
normalizing for volume, true peak normalization is a
more robust option as it oversample your signal to ensure that
there will be no clipping in the signal path. This tool measures volume
using DBTP, decibels true peak. We're more interested in the
loudness normalization options though. It's a bit of a
wordy name, but ITURBS17704 is a standard for
loudness developed by the International
Telecommunications Union, the same gang that
brought us classic hits like Rec.709 and Rec.2020. Loudness is measured in LKFS. That stands for Loudness
K-weighted relative to full scale. You might have heard
of this by another name, LUFS. That stands for Loudness
Units relative to full scale. It's essentially the same thing. This mode allows me to select a
target volume level and a target loudness level. Particularly when
working in broadcast, your client will have
specific delivery standards that you can enter here. For example, companies like Netflix, Disney, Spotify,
YouTube have published standards that they expect
videos to be delivered in. Fortunately for us, these
standards are already built into Resolve. I'm going to use the YouTube preset. You can see its
target volume and loudness levels are already
set to YouTube's published standards. Let's see how audio
normalization affects my delivery. I'm going to make
six deliveries, a normal mix, a delivery where the
audio has been mixed too loud, and a delivery where the
audio has been mixed too quiet. I'll then repeat each
delivery with audio normalization enabled. Let's check it out. And for that, I'm
going to pop my headphones on. To demonstrate the results, I've
imported each delivery into this timeline and will solo tracks to
demonstrate the differences. For reference, here is the normal mix. My mom still tells the story now
when she's doing her residency in New York. By the way, to solo a
new channel and remove the solo from the
previous channel in just one click, hold down on option or alt
while clicking on the solo button. New York City could always identify the ballerinas, could see
them on the subway, could always... The normalized version of the normal
mix has had a small amount of gain applied so that it hits
the target loudness level. Now, next, the quiet mix. I can hardly hear it. And now, the
normalized version of the quiet mix. The normalized version of the quiet
mix has had large amounts of gain applied so that again, it hits
the target loudness levels. And finally,
here's the distorted loud mix. Because she did. The genre of ballet is far
broader than people initially might think. It sounds terrible. And now, for the normalized version. Contemporary ballet,
there's less of a two-tone. It's about the physicality
and the athleticism of classical. The normalized
version has been attenuated so that it hits
the target loudness levels. The result is that all the
normalized deliveries match each other. So, despite my best
efforts, my delivery sounds great. Please note though,
best practice is to always manually master your
mix to the target standard. In this instance,
normalization was able to recover the quiet and
loud signals without loss. But it's possible that certain
operations can damage your signal irreparably. Before we finish, let's just
talk about what normalization is not. Normalization will
not help you mix your edit and normalization
will not manage dynamic range. It's still up to you to set the volume of different elements in
your mix relative to each other. It's also still up
to you to use tools like compressors to reduce
the amount of dynamic range and things like dialogue. So, normalization is
not the same as a limiter. A limiter is
basically a really aggressive compressor that again
affects the dynamic range of your mix. And again, it's still up to you to set up limiters to address
issues like errant peaks. Remember,
normalization applies a fixed amount of gain or
attenuation to your entire recording. So, a single loud sound will cause
the entire recording to get attenuated. While normalization on
export does not absolve you of the need to
properly mix and master your audio, you'll probably want to turn
this cool new feature on for most exports. Who doesn't want
more consistent deliveries? And if you're not working in a
specific broadcast or digital workflow, for most digital and online deliveries, the YouTube preset
is a good place to start. I love that this feature is
now available in the Deliver page. I hope you found this video
helpful, and if you did, please give it a like. Subscribe so you don't miss
future videos, and thank you for watching.
