Do you ever see one of your news stories or promos go out and realize loudness wasn’t checked or corrected properly?
Even if you have hardware-based audio compliance sitting on your output, issues with loudness can still arise. Watch this webinar replay “Loudness for News and Promos,” from Emotion Systems to discover how to implement more reliable and consistent loudness monitoring and correction.
Rich Hajdu:
Welcome to the webinar on loudness for news and promos. It’s a big issue because a lot of people do news segments and the audio goes out to the playout server and it’s not really loudness controlled.
I’m Rich Hajdu with Media Technology Group. MC Patel, the CEO of Emotion Systems, is going to be doing most of the presentation. We invite you to use the chat to send us questions and we’ll answer them.
MC, the issue is that loudness for news and promos is not regulated in many instances because people don’t have time to check the loudness when the news clip is edited. So it goes to the playout server and then it goes directly to playout. When it’s time goes through a playlist and all that. So there is a loudness corrector at the output of the station. Why isn’t that enough?
MC Patel:
The best thing is for me to quickly explain loudness. Now, most people will be familiar with it, but I’ll explain it very shortly and succinctly, and that will help you understand why we do the file-based loudness correction. Historically, ever since television has been around, we have used peak-based measurement. What we basically do is say, “What is the loudest or the highest peak in television?” That’s integrated in different countries at different time constants. In the US you guys use VU. In Europe, we use PPMs, but fundamentally what we say is that the sound shall not exceed a certain peak. In the old days, the reason for that was that in NTSC, the audio sat right at the end of the chroma subcarrier, so if the sound got too loud, it interfered in the transmitter and distorted the color.
That’s where the peaking came about. As you know, in the seventies, people discovered that whilst you weren’t supposed to exceed a peak, if you stayed very close to it, your commercials could be very, very loud, which is how the loud commercials came about. Over the years, people got fed up with it and said, “we want some balance in the content, we want dynamics, but we also want to make sure there aren’t sustained bursts of loudness.” So the new standard called “program loudness” came about — in America it’s called the CALM Act, but the standard is ATSC A/85. It said that the average level of a piece of content may not exceed a certain amount in LUFS. It’s minus 24 in the case of the US, and then the second parameter that’s important is the true peak. So the true peak may not exceed minus one or minus two, which varies depending on countries and standards.
When people used to mix to peaks, it was easy. They could see a peak on their PPM meters, and they came along and said, “Yeah, I’ve got my audio right.” When you have to measure to average level, that’s kind of hard because you don’t know what the average is till you get to the end. So when all this came about, there were a bunch of hardware correctors that you could put on the output of your master control and say, “All the content that I’ve been producing for all these years is great. And the hardware corrector will take care of the loudness.” Now, remembering that the average is what matters, the hardware corrector doesn’t know when a program has begun and when a program has ended.
So the hardware corrector is saying, “Over the last few seconds, whatever I’ve seen, I’ll keep it very close to minus 24.” So if you have a few seconds of silence, it says, “Oh, dear, my average is going to fall below minus 24.” It will raise the gain of that. And if you have a lot of very loud noises, it lowers it. Now that’s modulating the audio, and that can cause problems.
If we go back to the promos and the news, the important thing that Rich said is that in a promo, in a local edit suite, you may not have the loudness measurement capability, or that people still mix with their ears for promos or anything. And one of the nice things is if you mix with your ears and you’ve been trained, you will naturally mix to 24, or in Europe we think it should be 23, or thereabouts. So the mix will be good.
And that’s the other important thing: If you spend time creating a mix, you don’t want to destroy it in the correction process. For news, unfortunately, as we know, the news comes in, you’ve got some background noise. If you’re by a freeway, it will be really noisy background. If you have a quiet area, it could be really quiet. Then you have the people talking. And you’re not sure what the levels are going to be, when it comes there and it gets edited. And as Rich so rightly said, there isn’t time to take care of that. So what you’re relying on is the hardware processor to take care of it. Am I’m making sense so far?
Rich Hajdu:
Yeah. So the hardware processor isn’t really set up to do that. The question is, I’m editing news and it’s fast-paced, I don’t have time to manually intervene and check audio levels and all that. What’s an elegant solution to do that, that is settable, repeatable, and reliable? What’s the solution?
MC Patel:
There are a number of approaches. You can use a plugin in the edit suite, but as we already said, we don’t want to do that because the focus should be on getting the right piece of the video there and out. When we approached this problem, our initial focus was commercials. A lot of people came up and said, “Hey, we don’t want to be fined.” And the creative mix was important. A lot of the post houses said, “We don’t touch the audio because if we mess up the audio, our clients will get upset.” Bearing in mind that what we’re interested in is the average, what we say is we measure the average. Now in a file, you know the beginning, you know the end. So that would be applicable to a news clip or a promo clip, right? So what we do is we do two passes.
The first pass is we measure, then we say, “Oh, it should be minus 24. I am at minus 26.” So what we say is, “If you apply gain across the whole clip or two DBs, you’ll get to minus 24.” It’s a very simple algorithm. It’s volume control. Now, if you lifted the audio, then you may have a peak error because your peak’s not supposed to go above minus one. So what we do when we’re measuring it, we identify where the peaks are, we create a gate around it, and then the software attenuates inside the gate to make sure it doesn’t exceed the minus one. And then it does a little mix in, mix out with the main content. So it’s a very simple concept.
So you ask, “How do I do this?” Well, once we’ve established the parameters that the correction should be at minus 24, and the peak should not exceed minus two or minus three, different countries have slightly different standards, you set those up and then the software just takes care of it. You do a measurement pass, you do the correction pass.
In order to do this quickly, it doesn’t sit in the edit suite. The product can sit in there, but what we’ve got is a watch folder, or a hot folder as you might call them. We say the output of the edit suite gets posted into the hot folder, our product is looking at the hot folder. As soon as it receives the file, it measures it, it corrects it, and it puts the corrected file into a folder of your choice. So that could be your transmission server folder.
Rich Hajdu:
And how long does that take?
MC Patel:
Okay, so a 30-second file. Typically the software runs at three to five times faster than real time. So you’re talking if it’s five times, six seconds.
Rich Hajdu:
Right.
MC Patel:
And it’ll take us slightly longer because we watch the file to make sure that it stopped growing because when you’re copying a file into the watch folder, but other than that, it’s faster than real time. And depending on your setup, you can do this. So in the newsroom environment, it’s entirely feasible to make this happen and not worry about the audio side.
Rich Hajdu:
Yeah, because those clips are typically only 20 seconds, 30 seconds, a minute. In the US, if it’s a two minute clip, that’s a long clip. So that’s not going to add much time to it. Right?
MC Patel:
And you would have some time for the promos. So you could either do them manually, or again, you can have another watch folder or hot folder and do it that way.
Rich Hajdu:
So, okay, I’m going to do this. What is the actual physical implementation? In other words, I need a PC… what do I need from a hardware and software standpoint?
MC Patel:
The software’s cross-platform, first of all, and that’s important. So we can run on a Mac, on a Linux or on Windows. Depending, for example, it can be on the same PC or workstation as your edit software, the watch folder may be somewhere, but this can run on there. So you basically say, “Oh, we’re a Mac place. Can you run on Mac?” Yes, you can run on Mac. And it’s a standalone piece of software. It’s very easy to install. The installer comes with it. You download it from our website and it guides you through the installation.
We let people try this out, because if people have a concern that this may not be fast enough, or we’re not sure about the quality, what we say is, “Download it from our website and play with it.” If you need a little bit of help, we’re happy to help you set it up and so on. The correction profile for the US is a part of a library of profiles that are supplied with the application. So you don’t even need to set that up. You just select it and you’re done.
Rich Hajdu:
You talked about running your Emotion Eff software on the same PC that’s used for editing, whether it’s Adobe, Grass Valley, Avid, whatever it is. What if I have eight edit stations? Would I run that software on each one? Or would the watch folder be located on a central PC somewhere else?
MC Patel:
We have clients who do both. Obviously it costs money. If you’re doing a news spot, you would be doing half a dozen spots or whatever, right? So there’s no reason for everyone to have one of these. We’d love them to have it, but from a practical point of view, you can have a dedicated workstation and have a watch folder in there. And that takes care of it.
Now the computer doesn’t need a very beefy spec either because we’re only processing audio. Now we do read the file with the video in it, and we do write out a file with video in it. We do all the extraction, everything inside the software. So if you have it on a fast network, the network’s important, but in the lab, we have $500 PCs that take care of this.
Rich Hajdu:
So it’s a simple setup. Everybody in news is editing fast, they’re doing this and that. Once you set it up and it’s working, it’s a set-it-and-forget-it type of situation. Is that correct?
MC Patel:
Yes, it is. To give you some idea of the robustness of the product, we have two products really. The focus today is on what we call our desktop product, Eff, and we have another protocol, the Engine, that’s running pretty much identical code, but it’s designed to offer really, really high throughput. So we have clients processing 10,000 hours a month of content for retransmission. So all the episodic stuff, the movies and things that go out. If you have a playout center that says, “I have 50, 60, 100 channels, and all the content needs to be loudness compliant,” we have that going through the system. And that works day in, day out. It’s a very stable piece of software, and you wouldn’t have to have concerns about that.
Rich Hajdu:
If I install this software, I downloaded it, I installed it. Are you there to help me through the installation, answer questions, all of that?
MC Patel:
Absolutely. Now, whilst we are in England, if you give us team rear access, we can take care of the stuff. We can do online training in the same way. Iain, my partner, does all the support. This morning Red Bee France was doing a software upgrade, which they did themselves. They just wanted a bit of housekeeping check, so they gave our team rear access. Those guys are doing 2,000 hours a month. We have a call tomorrow in India where they want to set up the same idea. They are a sports channel, or they have lots of stuff, but they’re using us to correct the commercials that they put in, in between the sports. Obviously there isn’t a lot of live sports at the moment, but all the commercials and promos there go through this. So we’re helping with their setup because they have instances where they’re trying to deal with a 5.1 and Dolby E for transmission, which may not be the case in the US.
Rich Hajdu:
Right. And what kind of analytics are available?
MC Patel:
When we did this automated piece of software, people said, “How do I know what you’ve done? And give me some feedback.” If you’re doing it without the watch folder, if you’re doing a one-in, one-out, which I can show you in a second, the software draws a graph for you, and it gives you all the measurements. If you are doing it within the watch folder, you could set it to give you a PDF report. And within the report, it will tell you whether the file passed or failed, what it measured, what the graphs were, and so on. It’s very comprehensive in that. Now in your promos and things, if you have bars and tone, we can literally detect the tone and ignore it. So there’s a fair degree of flexibility.
Rich Hajdu:
Once the system is easy to install, it doesn’t require a lot of hardware. It’ll work on a network, doesn’t require a costly PC. It runs in the background, and it has analytics and it can be expanded. This is an important point. If you have another five edit bays, or you want to do more promos, or if you got into production, you could upgrade from Eff to Engine, so it’s a scalable system.
MC Patel:
Yes, it’s scalable.
I forgot to give you some background. We are Emotion Systems and we’ve been in this business for about 10 years. We used to sell some QC software whilst we were doing some other things. And a lot of our customers kept saying to us, “This QC stuff is really good. It tells me what’s wrong with a file, but there’s no opportunity, so what do I do about it?”
At the same time we’re doing this, people started talking to us about loudness. We thought, “Ah, this is interesting because there isn’t an easy solution of how to fix it loudness-wise.” So our first product was to build the measurement of loudness and correction of it. And the algorithm I described a few minutes ago is what we use. We’ve refined it a little bit, but fundamentally, that’s what we do. And then the company evolved to say, “Well, if you can do loudness and you’ve got the audio out, you can do more with it.” And that’s how the company’s grown. So the flexibility and the scalability came about, because once we did this, Dolby came to us and said, “What happens if the file’s got Dolby E? Can you measure the loudness in Dolby E and correct it?” And we licensed the stuff from them and made that.
And then Red Bee Media, about six or seven years ago, came to us and said, “We’re doing a playout center. We want to do loudness correction. We want to do channel mapping because our tracks are in the wrong place, etc.” So we built a system for them that could process 2,000 hours a month. But the desktop product has been around. It runs the same loudness code as the main Engine. So people often say we have the Eff, but now our needs have grown, and we’ll do an upgrade of the Eff into Engine. We do a trade-in, basically.
Rich Hajdu:
Right. And you have customers all over the world, and you have US customers also, correct?
MC Patel:
Yes, we do. We have some call letter stations that use Eff. We have a PBS station that does that. I won’t name them because I don’t have permission. There are very large content-repurposing houses, some of whom I can name, who, basically when Hollywood sells content to the rest of the world and locally, they want to create multiple versions of the same content. So we have companies like Vubiquity, Premiere Digital. Viacom’s a very big customer of ours. Viacom in New York has a massive system where a lot of their content goes through our system for loudness correction, pitch shifting, and a whole lot of other things.
Rich Hajdu:
MC, we don’t want to take too long here, but can we do a quick demo so we can see what this looks like?
MC Patel:
So what we have here is the product Eff. I’m going to show you the manual version. Let’s do a quick file measurement run.
We won’t go through the settings because I’ve already described most of it. All you’re doing is asking, “What is the correction I want?” So this is our CALM. I’ve got a Deluxe. I was doing a demo for them in Australia. Rich, when I did the demo for you. I have a Netflix profile, right? Now this is my demo system. When you get it, you actually get more than that, but we want to do CALM. I select a source. So this is a file already corrected.
Rich Hajdu:
Okay. Yeah. We just need to see the basics of what Eff does.
MC Patel:
So there you go. So what it’s done is it says the audio duration was 25 seconds at the top. And it measured it 19 times faster than real time. The analysis time was 1.3 seconds. Now this is a corrected file. So it says this group does not require any correction.
If it did require correction, I just press the correct button there and we’re done. And you can see that I’ve got a graph which says I have silence for the first seven seconds. And then the program loudness goes up like that. I can look at the true peak. It shows me where my true peaks are, and I can look at the log of these as well. So I can turn the logs on and it gives me a list where my peaks are.
Rich Hajdu:
So in a news editing situation, would the files that have been corrected be there?
MC Patel:
We’ve got a source and we can say, “Where do I want to put the destination?” Right. So I select the destination. Now I’ve actually just shown a corrected file, but it’s really as straightforward as that.
Rich Hajdu:
And so when I set that up and then as long as the destination doesn’t change, I set this up once.
MC Patel:
Correct. So I am doing this manually up here. That’s the manual operation. There is a watch folder operation. If you select the watch folder, then you can post files into a watch folder right from the system. Have I covered enough, Rich?
Rich Hajdu:
Yeah. I think you have. The point is it’s not a complicated, complex setup that takes days to set up and days to learn. Somebody gets this file and you guys help them. They can be on the road right away.
MC Patel:
The idea here is, this is an audio product that is designed for a video environment. So we assume that if the person doesn’t understand video, we’d take care of it.
You need a bit of trust. So we want you to try it out and get your audio people and the golden ears to listen to it and say, “Is this okay?” And then you’re good to go.
Rich Hajdu:
That’s the key, because an audiophile is different from somebody who’s concentrated on news. And really audio is just a secondary element that happens.
MC Patel:
That’s correct. So what we’re saying is, you focus on the creative stuff, do your job, and leave us to do the compliance.
Rich Hajdu:
Right. And they can download this file from your website, or they can go to my website and access the download and they can contact us. And it’s that easy. Just download it, try it out, and see if it works. We don’t want to get into pricing exactly. But let’s say I’ve got eight edit suites and this is not a $100,000 system, right?
MC Patel:
No, it’s not. It’s a few thousand dollars, depending on the configuration.
Rich Hajdu:
It’s in that price range. So it’s not going to be a budget buster either. And it is going to give somebody the continuity of audio without the highs and the lows and without all the other things that interfere with the viewer response.
MC Patel:
That’s correct. So we have some questions coming along. Cristian asked, what about live transmission? Basically, this product is not designed for live transmission. This product is really designed for file-based content. As I mentioned earlier for live content, we don’t know where the beginning of the program is and the end of the program. So we’ve left that part alone. We see so much content that is file-based that we saw an opportunity to make a product specifically for file-based content.
Rich Hajdu:
Cindy has posted in the chat box how you can download a demo version. And again, you can contact either one of us, me in the States and MC in the rest of the world, and we’ll be glad to answer any questions. Try the demo and make sure that you feel comfortable with the solution to an ongoing problem.
MC Patel:
We have a question from Europe, which is that you may want to preserve the archive as was originally mixed and then that archive may get distributed a number of times. A lot of our customers will do, I wouldn’t necessarily call it the correction, but the compliance to whatever the client needs. So for example, if you have something in your archive that has… I know we’re digressing slightly from the news and promo…
Rich Hajdu:
That’s okay. We’ve covered that.
MC Patel:
If you have something from the archive and you wanted to send it to Netflix, for example, Netflix has a completely different standard from the broadcasters. So this product is very often used for producing content for Netflix or for Amazon Alexa.
Yesterday I was talking to some people who said, “We have a completely different audio requirement for Alexa. If I already have a hardware processor, why do I need this?” And as I mentioned earlier, what’s happening to most hardware processors is they’re programmed to make sure that the long form content is compliant. So the time constant that it reacts to are over long term changes. And that’s also true for live. So for live, you have quiet moments, really loud moments, quiet moments, really loud moments. And it could be a three-hour, well, if you watch cricket, it’s an eight-hour show, but like a baseball game, I guess it’s many hours, right?
So the processors are really set for that. When you have a 30-second, 45-second promo, you really want the promo to be bang on in the range so the hardware doesn’t try to overcorrect it or undercorrect it.
Another question is API integration. The product that I briefly showed you, Eff, is a desktop product. It’s got a little bit of automation in it, but it doesn’t really go beyond that. If you need API and so on, then the product to look at is the Engine which has 16 different audio processing modules. It can process up to eight files at a time. It’s a very, very comprehensive system designed for scaling and automation, and we have a REST API for it. And what happens with a REST API is that you could basically say, “Take this file, apply this workflow to it and send it to this position or this destination.”
It’s as simple as that. Once you’re doing it, you can do the status of the file in terms of how much you’ve processed it and so on. So there’s a question on latency and delay. In this instance, there is no delay or latency because it’s software — what we’re doing is we’re taking the audio out of the file. We’re measuring it, we’re correcting and putting it exactly back where it was. So there’s no video, audio time constraints. In hardware, there may be a different approach to that.
Rich Hajdu:
In the chat box, there’s also an area where you can download the Loudness Factbook, which is really, really excellent. I’m not an audio expert, I’ve spent my time in television. And the Loudness Factbook is really good because it gives you a great primer on loudness in all its attributes and all of that.
MC Patel:
We have a question which says, “If someone is doing a poor job with the audio track, can you fix it?” Now what we have to ask ourselves is, what is a poor job? I often describe this as, if you think about music, I could say the mix on this music is bad, but the artist created it. So we try and preserve that. But if what you call bad is the audio levels are too low, I can’t hear the dialogue, etc., when you normalize the file, because it’s applying that overall gain, you can actually do a few things to it. In fact, that’s what we’re saying for the newsroom. Generally, the problem is people do know how to talk, but sometimes they have to shout because they have to get above the ambient level. And we can normalize that so that is better, but it is literally a gain control.
I have a question from Bogdan. Does it correct loudness on Dolby E streams? And the answer is if you buy the option, it does. Bogdan, I assume you’re in Europe at the moment because we use Dolby E quite a lot. The Red Bee example I use for playout does have Dolby E. So some of the files come transmission-ready. They already have Dolby E and if you want to say, “Is the loudness in that stream correct?”, we decode it, measure it, correct it, and resend it.
This is also applicable in the US if you have to deliver content for someone like Direct TV. You may deliver it as transport stream with Dolby digitally encoded, or MP3 in case of Comcast. And then you say, “I’ve sent it, but is it correct after the encoding process?”
We have done a system like this for a large content creator, where we take the file and we decode the Dolby Digital, loudness-measure it, correct it, recode it, and give it back. Now the important thing here is if you’re the guy who has to send the file out, you may not have the skills to do this and you sure as hell won’t have the kit to do it. So this is like, “I have to send it out and now to make sure it’s okay.” We provide that capability.
Rich, have I covered everything for the news?
Rich Hajdu:
Yes, MC, very productive. Because again, we needed to know about the simplicity, how it works, the repeatability, the reliability, and basically what we’ve discovered in this webinar is that it’s a simple system to install. It’s a simple system to use. It’s repeatable. Set it and forget it, and it’s not too expensive. So I think that’s everything we need to know. If anybody has any questions, they can address MC or me and we’ll be glad to answer them.
MC Patel:
Once again, as you mentioned the Loudness Factbook is something we wrote to help people understand it with all the issues. There is a discussion in there about the hardware versus software, where it’s applicable, why you should do one or the other. We are always happy to discuss specific needs. Just drop us an email on . And you can go onto the website at emotion-systems.com. Every single page on our website has a big button that says “Download a trial” and click on it and you can download the software. We’d love to work with you.
Rich Hajdu:
Great. If that’s it, then we’ll close it for now. And thanks for everyone’s participation. It’s been very enjoyable.
MC Patel:
Yeah. Thank you for your time, gentlemen. Thank you.
