Fairlight Loudness Scan

Fairlight is a great tool for audio mixing, and the Loudness Meters work very well to help us get the proper levels for our projects.

But at the moment, we have to play a project in real time, in it's entirety, to get a proper reading from those meters. (Speeding up playback will not give the correct result.) If changes are made, that's another full program playback. More changes, more playback. Ugh!

We desperately need a way to scan through a full program as fast the the system can handle, while still giving us an accurate Loudness Meter readout.

+1000000

That would indeed be great. I am not aware of any other loundess meter VST plugin that can do that though so there must be some technical issue with that. Every plugin seems to require you to play the whole video from sart to finish in real time in order to give an accurate loudness value.

That would indeed be great. I am not aware of any other loundess meter VST plugin that can do that though so there must be some technical issue with that. Every plugin seems to require you to play the whole video from sart to finish in real time in order to give an accurate loudness value.

E.g. Waves WLM Plus:
https://www.waves.com/plugins/wlm-loudness-meter#image

I am not aware of any other loundess meter VST plugin that can do [this]

I'm not a big fan of plug-ins. Too many potential problems. So I do hope that if any such ever become available, BMD will not take that as 'solved'.

This feature needs to exist natively.

Audition can do it. Fairlight needs to as well. I hope before 16.1 goes Gold.

+1

A manually triggered faster than real time loudness analysis function would be a hugely useful feature.

I use the lufs meter plugin and bounce the track, and it gives accurate results.

Again, plug-ins are not a viable answer here. I think this capability needs to exist natively.

Again, plug-ins are not a viable answer here. I think this capability needs to exist natively.

I agree it should exist natively, but it is possible today to work around the issue until BM implement something.

Personally, I do not think it's a good idea for BMD to be a judge and a party?
How to display in Fairlight Page a loundness on the one hand and
to display a different result via an OFX on the other hand?
Contradictory?

In addition to the Loudness analysis, the scan also needs to detect True Peak.
Then it will be perfectly.

In addition to the Loudness analysis, the scan also needs to detect True Peak.

Absolutely. (And show us visual indicators of where in the program those True Peaks are, so they're easy to find and fix.)

Hi, I’m new here, I’m an amateur and I support this requirement.
Loudness analysis as fast as my CPU/GPU can do would be a great timesaver.
But the numerical values are not enough (IMHO). I would love to see the curves in the loudness history to be able to find the problematic points.

I am not aware of any other loundess meter VST plugin that can do that though so there must be some technical issue with that.

My hunch is that the VST protocol doesn't give a plugin the necessary 'permissions' to trigger a faster-than-realtime playback needed for that type of measurement.

Personally, I do not think it's a good idea for BMD to be a judge and a party?
How to display in Fairlight Page a loundness on the one hand and
to display a different result via an OFX on the other hand?
Contradictory?

I'm assuming everyone is talking about the same standard though so the results should be identical. Nothing to worry about.

If changes are made, that's another full program playback. More changes, more playback. Ugh!

Well, at least with US standards there's really no need to re-measure if the first pass hits -24LUFS on the nose. It's +/- 2LU so any changes made often aren't big enough to move that value beyond 2 units, in my experience.

Perhaps a better way to put it is that the bigger or more numerous the changes are the more time should be allotted for it and that includes measuring. I mean if I have to do major surgery to a 1hr show then I won't worry about the time it takes to either watch the entire program beginning to end in realtime to verify the changes (a good idea, no?) or do a faster-than-realtime export and then offline analysis of the resulting file. And if the changes aren't big or numerous then it either doesn't take long to measure in realtime or render/analyze using a plugin isn't a big deal.

For reference I've just finished a series in the US and I had plenty of notes on each episode, and I just printed stuff. The first mix hit -24LUFS on the nose and nothing I did subsequently moved that target much at all. Probably not more than +/- 0.7LU.

But the numerical values are not enough (IMHO). I would love to see the curves in the loudness history to be able to find the problematic points.

I agree that this is useful.

For anyone using plugins for metering/analysis though some do contain a history 'curve'. iZotope Insight is one.

Jim,

that might not be feasible. LU/LUFS is based on a weighted real time measurement with multiple measurement windows, including short, medium and full program, along with True Peak.

Doing a faster than real time measurement on a timeline would probably mess with the weighting at very least (it's frequency based, loosely after Fletcher-Munson).

There are a number of offline tools available to measure loudness, from Dolby, NuGen Audio, RTW, Pinguin Software, along with lesser-known options. To do an offline measurement you need a final deliverable, i.e. a finished render of your project.

If speed is the issue, it may be that a faster than real time delivery (cutting the video/effects to the bare minimum or doing an audio-only delivery) to a file will then allow you to use any of the aforementioned offline measurement tools to verify you're within spec. I don't know of a faster way to do it.

I find in general that if you have your studio monitoring environment properly calibrated and keep an eye on your levels as you work through a project, in all but a few rare cases you'll hit the -23LUFS target pretty squarely. Getting your ears habituated to the "right" levels over time in the studio will help dial in the proper overall and dialogue loudness pretty quickly, and the meters serve as a useful confirmation that you're in the green zone.

Hope this helps.

I don't know of a faster way to do it.

Exactly. Providing that 'offline' workflow directly within the UI could be a great feature.

For example, a 'Generate Loudness Print' (or some such Fairlight menu option) could initiate a modal operation (ie app presents a progress window preventing other operations) during which an audio only mixdown would be automatically rendered to a temp file, scanned/processed (faster than realtime), and the loudness data displayed, as its generated, as a loudness 'track' on the timeline. To my mind that would effectively create a linear snapshot (ie valid at time time of creation) of the loudness graph of the full sequence (or just a marked section). It could be displayed in 'green' whilst valid, then change to 'red' as soon as it was invalidated (ie as soon as any part of the audio in the timeline was edited).

I like to count myself as someone with well a calibrated monitoring environment who doesn't usually have many problems with mixes... but a tool/option as described could still certainly be useful.

That's more or less the way Nuendo does it. There's a loudness track that can calculate in realtime or faster. If it's done faster then it's as you say, it renders the mix and analyses it but only displays the resulting loudness line.

I will say this though; there have been some people complaining about how long that takes. It is what it is of course, and it can only be done as fast as the computer can render a mix faster-than-realtime, but to many it's long enough a time to wait that for 1hr shows (or films) it's annoying.

But yeah, that's one way of doing it.

Just to be the lone voice of dissent in the wilderness, while I appreciate the peace of mind and verification an offline total loudness measurement can provide, if you're keeping an eye on your levels as you work, you will in most cases be fine. In those few cases where you're +/- 1-2dB off, you can adjust the master fader accordingly and re-render.

Really, keep things at the proper levels in the mix and you're good, and are building good habits as well as training your ears to recognize how things "should" sound at the proper levels.

One way to help ensure this is to work to calibrated levels in the studio and don't touch your monitor gain control once you've set it. Things tend to fall into place then.

Just offering a different perspective.

Just to be the lone voice of dissent

I pretty much said the same in the other thread...

Really, keep things at the proper levels in the mix and you're good, and are building good habits...

Yep, no one here is suggesting otherwise. That's just part of the job and what all do already. It doesn't really have any bearing on this feature request.

In those few cases where you're +/- 1-2dB off, you can adjust the master fader accordingly and re-render.

Correct. And this is what people do already... but to get that total loudness reading that shows that either you're already spot on, or that you need to maybe adjust the master fader very slightly, you have to play through, start to finish, in real time. The point to this feature request is that it would be more efficient if that were a 'faster than real time' process. At least that's my perspective.

to get that total loudness reading that shows that either you're already spot on, or that you need to maybe adjust the master fader very slightly, you have to play through, start to finish, in real time. The point to this feature request is that it would be more efficient if that were a 'faster than real time' process. At least that's my perspective.

I think people agree that it would be a nice feature to have. But what his (and my) point was was that when we work - even when we start and stop - we end up with those short- to medium-long sections and thus the entire content on target, and we also typically watch the content beginning to end at least once before delivering anything to other people.

As I mentioned here or in the other thread on the topic I've just delivered the last episode in a series and not once did I use the faster "offline" measurement for it. As I'm editing I keep track of rough levels. As I'm mixing I'm keeping track of them. Even though I might run sections as short as 10 seconds they're never going to be wildly off from other sections. And then I watch the whole thing while taking notes on stuff before laying off the mix to the client, and at the end of that watch-through I have the measurement. I'm in the US so it's +/-2LU and I'm always within about +/-0.8. So I won't even have to adjust anything for that target to be met.

So it just doesn't really seem like a big issue if your workflow is good.

there's really no need to re-measure if the first pass hits -24LUFS on the nose.

Agreed.

But when it doesn't, a new real time measurement needs to be taken after each round of changes. That's a real time suck which can be solved by a faster-than-real-time scan.

that might not be feasible.

The feature exists in Adobe Audition, so...

The point to this feature request is that it would be more efficient if that were a 'faster than real time' process. At least that's my perspective.

Exactly right.

there's really no need to re-measure if the first pass hits -24LUFS on the nose.

Agreed.

But when it doesn't, a new real time measurement needs to be taken after each round of changes. That's a real time suck which can be solved by a faster-than-real-time scan.

I understand what you're saying, I just can't even remember the last time I wasn't on-target. It basically sounds like a problem with the audio-post workflow (or engineer).

I agree with Mattias.

This is an audio engineer problem, not a software problem.

IMHO.

I think people agree that it would be a nice feature to have. But what his (and my) point was was that when we work - even when we start and stop - we end up with those short- to medium-long sections and thus the entire content on target, and we also typically watch the content beginning to end at least once before delivering anything to other people.

As I mentioned here or in the other thread on the topic I've just delivered the last episode in a series and not once did I use the faster "offline" measurement for it. As I'm editing I keep track of rough levels. As I'm mixing I'm keeping track of them. Even though I might run sections as short as 10 seconds they're never going to be wildly off from other sections. And then I watch the whole thing while taking notes on stuff before laying off the mix to the client, and at the end of that watch-through I have the measurement. I'm in the US so it's +/-2LU and I'm always within about +/-0.8. So I won't even have to adjust anything for that target to be met.

So it just doesn't really seem like a big issue if your workflow is good.

Your view of working with audio in video focuses only on the phase of production of audio-visual content, but you need to measure audio, especially the finished material. E.g. when preparing content for television broadcasts to different parts of the world, where standards for sound are different, for cinemas, VoD platforms, etc.

Meter capabilities such as Waves WLM Plus Loudness Meter or iZotope Insight are used for this purpose and integrated into DaVinci / Fairlight would be very beneficial.

Your view of working with audio in video focuses only on the phase of production of audio-visual content, but you need to measure audio, especially the finished material. E.g. when preparing content for television broadcasts to different parts of the world, where standards for sound are different, for cinemas, VoD platforms, etc.

I don't understand your point above. You say I focus on audio-visual content, then say "but" and that I need to measure audio when prepping for content for television. Content for television is audio-visual, so why are you saying "but"?

Yes, standards are different, but my point is that when you're doing your work in a calibrated room (one way or another) with a meter available - as you go along editing, premixing and mixing you will see where your level is and typically hit your target within the acceptable range.

And if you're talking about global deliverables for TV for example then you can pick the tighter European spec of -23LKFS +/-1LU which fits "inside" the US spec for example. Obviously if there's a desire to adjust True Peak differently for different networks that's an issue, but none of this changes anything.

Look at it this way: If you're mixing the same content for cinema as you are for TV then it doesn't change the fact that you either just brute-force limit and boost/cut the loudness of a cinema mix or re-mix it for that other medium. And once you're done with your cinema mix you have your measurement and don't need to run it faster-than-realtime again really. Mixing for cinema is such a big endeavor relatively speaking that of course you're going to have a play-through beginning to end at the end of it which is when you get your measurement. And then you either use the brute-force method based on those numbers, or you re-mix in which case we're back to what I was saying earlier.

Meter capabilities such as Waves WLM Plus Loudness Meter or iZotope Insight are used for this purpose and integrated into DaVinci / Fairlight would be very beneficial.

Yeah, I know those are used for those purposes because I use them. But Insight doesn't allow for an offline measurement of the mixer's output, and neither does WLM if I remember correctly. They can analyze files (clips) offline, but that's a different thing.

I already agreed that it would be "nice" or "convenient" to allow for faster-than-realtime. All I'm saying is that it isn't really "desperately needed" if your workflow is decent.

PS: Depending on the content and the complexity of the mix we might save slightly less time than we think... also depending on the computer of course. But as a somewhat realworld-esque hypothetical example: If you have a 40min TV show that's a decent size mix, let's say not a fully-filled M&E and not fiction but more of a TV doc or something reasonable, that may on an octo core CPU desktop take about 10min to render. This means that the offline measurement takes 10min to get. Like I said, some people have complained about that. Of course a "worse" CPU takes longer. And if your workflow is such that every time you make a change you need to remeasure then that's 10 minutes every single time. And that's outside of actually rendering the content.

Well, 17 has made some inroads here. We can now scan for Peaks.

Well, 17 has made some inroads here. We can now scan for Peaks.

I read it in list of news and I am glad. I will try how it work in the future because I am still on DVR 16.

I tested it, it works well

You can use this too (it's very quick) when you're outside Resolve : https://www.loudnesspenalty.com/ (usefull for youtube uploads with their penalty meter)

or this one (it's instantaneous) : https://youlean.co/file-loudness-meter/

I tested it, it works well.

Does Resolve show where the peak(s) on the timeline is (are), or will it print the timecode(s)?

... or this one (it's instantaneous) : https://youlean.co/file-loudness-meter/

I know Youlean meter but I don't use it. Can Youlean create a peaks list offline like a Waves WML Plus plugin?

Does Resolve show where the peak(s) on the timeline is (are), or will it print the timecode(s)?

This analysis only shows what's in Maxwellx's screencap. (Like True Peak and LUFS.) It doesn't do anything else currently.

Version 17 has been officially released, and man what an incredible update!

BMD can't do everything we want in the "next version", so I wanted to bump this idea, which didn't make the cut for 17.

Here's a feature new in 17 that's on track. We can right click a clip and Analyze to get the following.

Analyze Audio.png (8.77 KiB) Viewed 7336 times

But it doesn't go far enough. What we need is to scan the entire timeline as fast as the machine can go (skipping all video for this one) and getting a final reading of the following.

Analyze Timeline.png (13.95 KiB) Viewed 7336 times

Right now the only way to get these much needed readings from a full timeline is real time playback.

That's not good enough.

Orban loudness meter will analyze wav or MP4. Free + good.

+1

The ability to do a quick loudness analysis of a clip is fine, but being able to run it on a track or a bus would be much more useful (for me)... preferably (for me) generating a faster than realtime loudness plot, just as currently possible with real time playback.

Orban requires us to export the file. The point of this request is to know the readings before export. (Natively, without the use of a plug-in.)

This would definitely be a useful feature. As it is, in order to detect my integrated loudness reading, I have to set my limiter on the master, take a decent guess at a good amount of gain, export my timeline as an audio file, and drop it in Youlean's standalone app because that's still way faster than letting the timeline play at 1.0x speed all the way through. The fact that it can analyze the loudness on a per-clip basis but not a per-timeline basis is a big miss, in my opinion.

This would definitely be a useful feature. As it is, in order to detect my integrated loudness reading, I have to set my limiter on the master, take a decent guess at a good amount of gain, export my timeline as an audio file, and drop it in Youlean's standalone app because that's still way faster than letting the timeline play at 1.0x speed all the way through. The fact that it can analyze the loudness on a per-clip basis but not a per-timeline basis is a big miss, in my opinion.

But first of all I don't see why you would need to do the above as a first step it seems. It reads as if you're actually not mixing your audio at all. If you are then as I said above you should have a workflow that includes hitting the ballpark of whatever loudness you need to hit.

Also, if you do a faster-than-realtime loudness analysis Fairlight has to create the mix during that process and then measure it. So since it takes that amount of time you might as well render the mix anyway, no? Because in that case best case scenario you're good. You have your mix and you're done. Worst case you adjust and re-export in which case all you've lost is the time it took to analyze the imported clip/mix (which should be way faster than analyzing the mix). By contrast if you don't do it that way then you would have to run the mix one time for the measurement and then if it's good do it one more time to get the resulting file. That actually takes longer if your mix is good.

since it takes that amount of time you might as well render the mix anyway, no?

In faster than real-time, yes.

As a file on the hard drive, no. It isn't needed.

since it takes that amount of time you might as well render the mix anyway, no?

In faster than real-time, yes.

As a file on the hard drive, no. It isn't needed.

I know. The thing I was pointing out was that you need to do the faster than real-time export to get that file on the drive in the first place, if we're talking about measuring mixes.

This would definitely be a useful feature. As it is, in order to detect my integrated loudness reading, I have to set my limiter on the master, take a decent guess at a good amount of gain, export my timeline as an audio file, and drop it in Youlean's standalone app because that's still way faster than letting the timeline play at 1.0x speed all the way through. The fact that it can analyze the loudness on a per-clip basis but not a per-timeline basis is a big miss, in my opinion.

But first of all I don't see why you would need to do the above as a first step it seems. It reads as if you're actually not mixing your audio at all. If you are then as I said above you should have a workflow that includes hitting the ballpark of whatever loudness you need to hit.

Also, if you do a faster-than-realtime loudness analysis Fairlight has to create the mix during that process and then measure it. So since it takes that amount of time you might as well render the mix anyway, no? Because in that case best case scenario you're good. You have your mix and you're done. Worst case you adjust and re-export in which case all you've lost is the time it took to analyze the imported clip/mix (which should be way faster than analyzing the mix). By contrast if you don't do it that way then you would have to run the mix one time for the measurement and then if it's good do it one more time to get the resulting file. That actually takes longer if your mix is good.

I think we're saying the same thing. I get my mix in the ballpark of where I want it to be (-14 LUFS for YouTube) but YouTube and other platforms measures on the integrated loudness, which takes the entire duration into account. So you never know if the integrated loudness reading is at (for example) -13 or -16 until you run an analysis on the whole mix front to back. So I render just the audio, and have YouLean Loudness Meter analyze it. If it comes back as -16 then I know I need to apply 2dB of gain to my limiter, then export another copy to make sure I didn't go over (because sometimes adding 2.0 dB changes the measured integrated loudness by 2.1 dB), analyze that, and if it's all good I can do the actual export. It would be easier if the analysis was built-in to Fairlight, like it is with clips.

I think we're saying the same thing. I get my mix in the ballpark of where I want it to be (-14 LUFS for YouTube) but YouTube and other platforms measures on the integrated loudness, which takes the entire duration into account. So you never know if the integrated loudness reading is at (for example) -13 or -16 until you run an analysis on the whole mix front to back. So I render just the audio, and have YouLean Loudness Meter analyze it. If it comes back as -16 then I know I need to apply 2dB of gain to my limiter, then export another copy to make sure I didn't go over (because sometimes adding 2.0 dB changes the measured integrated loudness by 2.1 dB), analyze that, and if it's all good I can do the actual export. It would be easier if the analysis was built-in to Fairlight, like it is with clips.

Understood. It would indeed be easier if you didn't have to import/measure the rendered file and instead just measure the mix without exporting/importing audio.

I still maintain though that just a review from beginning to end without stopping solves that, even though it's a realtime process. It's generally good practice.

As for YouTube specifically I'm not sure it matters much if you hit -13, -14 or -16 when delivering if the YouTube system automatically normalizes it. Unless the audio / mix is unwieldly their limiting in case of boosting levels should probably not do much harm. -16 is already pretty squashed in my opinion, so not sure how much worse it gets chopping off another two dB. Which all just to say that adding 2.0 and ending up with a 2.1 dB change shouldn't really matter. I mean, it's not like they'd reject the upload, right?

I think we're saying the same thing. I get my mix in the ballpark of where I want it to be (-14 LUFS for YouTube) but YouTube and other platforms measures on the integrated loudness, which takes the entire duration into account. So you never know if the integrated loudness reading is at (for example) -13 or -16 until you run an analysis on the whole mix front to back. So I render just the audio, and have YouLean Loudness Meter analyze it. If it comes back as -16 then I know I need to apply 2dB of gain to my limiter, then export another copy to make sure I didn't go over (because sometimes adding 2.0 dB changes the measured integrated loudness by 2.1 dB), analyze that, and if it's all good I can do the actual export. It would be easier if the analysis was built-in to Fairlight, like it is with clips.

Understood. It would indeed be easier if you didn't have to import/measure the rendered file and instead just measure the mix without exporting/importing audio.

I still maintain though that just a review from beginning to end without stopping solves that, even though it's a realtime process. It's generally good practice.

As for YouTube specifically I'm not sure it matters much if you hit -13, -14 or -16 when delivering if the YouTube system automatically normalizes it. Unless the audio / mix is unwieldly their limiting in case of boosting levels should probably not do much harm. -16 is already pretty squashed in my opinion, so not sure how much worse it gets chopping off another two dB. Which all just to say that adding 2.0 and ending up with a 2.1 dB change shouldn't really matter. I mean, it's not like they'd reject the upload, right?

I totally agree about it being good practice to listen to the mix to make sure everything is dialed in just right, but I'm often racing to meet deadlines so I optimize my time where I can.

The thing with youtube is that it will normalize down but not up. If you're too loud, it'll turn you down, but if you're quiet, it'll just let it be.

The other aspect of all of this is: with YouTube, to be frank, I'm not mixing for audiophiles with home theaters, I'm mixing for people watching on their iPhone so making sure they can hear the dialogue trumps all.

a review from beginning to end without stopping solves that, even though it's a realtime process. It's generally good practice.

It sucks when you have a three hour timeline.

a review from beginning to end without stopping solves that, even though it's a realtime process. It's generally good practice.

It sucks when you have a three hour timeline.

Or more. Much more.
For film distribution, I need to measure the sound for 8-10 feature films a day.

a review from beginning to end without stopping solves that, even though it's a realtime process. It's generally good practice.

It sucks when you have a three hour timeline.

Or more. Much more.
For film distribution, I need to measure the sound for 8-10 feature films a day.

Those 8-10 films are just stems you need to sum and measure or they're including the stereo mix files?

Those 8-10 films are just stems you need to sum and measure or they're including the stereo mix files?

These are complete films: Video + Audio, 5.1, 2.0 matrix, 2.0, language versions. That's different.

Movies are delivered to different countries with different sound standards on different platforms: Cinema (not all cinemas want DCP), television, VOD platforms. It's different too.

The work on the film does not end with the export of the cut work. He still has to get to the viewer. This is the purpose of film distribution, which sells, repairs, censoring for non-democratic countries, changes distribution logos, adds language versions, subtitles, captions, archives, etc. etc.

My personal record is 14 movies delivered per day.