Hello
I’m editing for a few different podcasts and I want each episode to be A) In range with mainstream podcasts such as NPR, WTF, or Conan Needs A Friend. and B) Consistently Uniform from 1 episode to the next episode. Another example, I edited some audio extracted from a video and the customer asked for the audio to be -17 LUFs … I believe. ( Forgive me for my loose audio jargon.) I just want to know how to use the audacity editing work-field like a ruler. I guess I’m wondering if I can select a slice of audio and analyze it to get a decibel, gain, or Luf Measurement. I’m sorry if this question is all over the place.
I would love a ruler, sonic yardstick, loudness caliper to make sure all the audio I edit goes out within a standard range among podcast on Spotify and Apple.
[u]Loudness Normalization[/u] can be used to set the loudness to -17dB LUFS. (Choose the Perceived Loudness option.)
Loudness Normalization doesn’t check for clipping so you MAY need to run the Limiter to bring-down the peaks after changing the gain. (The limiter will have very-little effect on the LUFS level.)
Unfortunately, there is no built-in way to measure LUFS but [u]dpMeter[/u] (free) works in Audacity.
And it’s not the worst idea to use all three of the Audiobook Tools, not leave any out.
Home microphones like to produce rumble and low pitch tones and trash in addition to the voice. Normally, nobody can hear them and nobody cares (and they’re expensive for the manufacturer to fix). But if your goal is consistent voice volume, you don’t want them in there because they can throw off the volume correction tools.
So this is the Audiobook Suite.
Your version of it can look like this.
Effect > Filter curve… > Manage > Factory Presets > : Low roll-off for speech > OK.
Effect > Loudness Normalization…: Normalize Perceived Loudness to -17LUFS > OK.
Effect > Limiter: Soft Limit, 0.00, 0.00, -1.0dB, 10.00, No > OK.
The soft limiter is handy because it keeps your voice out of overload and clipping damage and you can’t hear it working. In this exact instance, It’s set to catch your voice just before overload. The audiobook people have a slightly quieter standard.
Note there are no compressors and processors in there, so if you like to get theatrical and expressive, you’re going to create problems.
Also if you have high background noise (traffic, refrigerators, your computer), that’s going to be processed right along with your voice.
It should be possible to compare your works to commercial productions directly and if your volume is significantly off try changing the LUFS value. If your LUFS value has to be a certain number and the show volume is still off, that’s a problem.
No matter what you do, you can’t get away from having a studio (quiet, echo-free room).
There is an explainer Youtube artist who has the video parts exactly right. He’s beautiful to look at, but he sounds like a kid recording in a bathroom.
Since your customer asked for the audio to be at -17 LUFS,
please be aware that Audacity has some issues when using external plugins on both Mac and WIndows.
It sends the wrong level (too high) to the plugins, so any LUFS measurement will not be true.
(BTW RMS and peak readings on VU/PPM meters etc will also show wrong values).
See here:
and here:
Newer versions (which I don’t use), do have some sort of Loudness Normalization built-in,
but have never used it so can’t comment.
As @DVDdoug wrote, it does not check for clipping.
After loudness normalizing, ensure that the waveform peaks have a bit of space and don’t touch the top or bottom of the track. If they do touch (or go over), then use the Limiter effect to bring the peaks down to about -1 dB using the “Soft Limiter” setting (see: https://manual.audacityteam.org/man/limiter.html)
If audio is set to -17 LUFS, there is no way that it should be slamming into the extremes of the “track”. @JimmyFro, if it is, best double check everything again.
Below, a comparison of audio at -23, -17 and -10 LUFS, note the “headroom” in each case.
A user on Reddit, Chaos_Klaus, put the whole LUFS thing into perspective:
(Please note that it was written about 3 years ago, so best check current levels for these platforms).
There is so much confusion about this …
Quite a while ago, the EBU has given out a recommendation on how to measure loudness of program material.
It works pretty much like dB RMS just with a filter and a gate applied.
To make sure everybody is clear on how these values were measured, we use LU (loudness units) or
LUFS (loudness units full scale).
The entire program is measured to find the integrated loudness.
There were also recommendations on what you should aim for, the loudness targets,
first for broadcast (radio, television) and later for streaming services and mobile devices too.
The recommendation for streaming services is to have your program loudness at -16LUFS to -20LUFS.
iTunes is currently the only platform which does actually comply with these recommendations.
It normalizes at -16LUFS. Spotify used to normalize at -11LUFS, which is super loud.
If your mix is not that loud, iTunes would simply turn it down so the peaks would not clip.
Spotify on the other hand, does actually use peak limiting to get quieter masters to that level! (Insane, I know …)
So recently Spotify has lowered that level to -14LUFS which is a good thing. It’s still not really optimal. >
Beware of the term “dynamic range”. It’s often confused with the “crest factor”.
Dynamic range is the difference between the loudest and the quietest part of a song.
So while your verse might be playing along at -16LUFS, your chorus might be smashing in your face at -12LUFS.
Crest factor however is the difference between RMS level and peak level,
and that is what greatly benefits from the lower loudness targets.
Was that done with the built-in normalizer?
Is it possible for you to post a non normalized version? or is Mahler’s music still under some copyright?
Would love to test it and compare with the studio tools we use.
Classical & chamber music is notoriously difficult to broadcast due to it’s very dynamic nature.
Often, a bit of subtle compression before normalizing it brings it into spec.
Of course the cost is some slight loss of dynamic range, but I don’t think true aficionados
will be consuming a majority of their classical music from TV broadcasts, but rather from
vinyl, CD’s and other high quality streaming platforms like soundcloud that have the option to
leave levels as they are uploaded.
EDIT:
Even FM stations broadcasting this kind of music, tend to “cheat”.
Their studio output first goes via a processor that not only tweaks the EQ, but levels and compression as well,
all in realtime.
A very typical processor is the Orban Optimod range.
Buying a copy doesn’t give you the copyright (which is more like a publishing or distribution right). YouTube is tricky because sometimes the copyright holder allows it and they collect payments which come indirectly from the advertisers.
As it turns out, didn’t have to buy anything, we have a performance in our library.
Not sure if it’s the same performance that Steve has, but either way, it’s heck of a dynamic,
just look at the LRA.
Since it’s for broadcast, someone already normalized it to -23 LUFS.
Format was 48KHZ, 24 bit wav.
With that LRA, hate to have been the person, who knows what the “raw” was like.
I re-normalized it to -17 LUFS using Izotope Ozone 9 (in Reaper), it did do some gentle limiting,
as can be seen from the resulting LRA, it’s slightly down.
Since most users here will not have access to the tools I used, the “soft” limiter
in Audacity is a very good substitute.
Then again (and as Steve wrote), doubt very much that the average user will come across audio
as dynamic as this.
The end section of the waveform, is applause.
Either there was an audience mic/s, or, they really liked it.
Great stuff! I’m gobbling it all up and taking notes.
Now, what about leveling? My method currently is to roll lows off via filter curve - Soft Clip, Amplify to ceiling, compress using the make-up gain checked box. This is the only sure-fire way to raise those valleys and make the recording optimally consistent. As long as I do a moderate noise removal pass after mixing down, the noise is a non factor. What’s the right way to do what I’m attempting and slopilly pulling off?
