The VideoVerse
The VideoVerse
TVV EP 28 - Romain Bouqueau: Championing Open-Source in the Streaming Industry
In this episode, Romain Bouqueau, CEO and Founder of Motion Spell gives us a deep look into the contributions of the open-source community in the world of video streaming. Romain also shares his insights into how open-source works, how GPAC/Motion Spell has remained ahead of the curve with its focus on R&D, and how open-source and commercial entities can collaborate effectively to fuel innovation in streaming.
Welcome to "The VideoVerse.'
Zoe: Hi everybody, this is Zoe. Welcome back to "The VideoVerse" podcast. So this time, I have Krishna, our team member who is based in Bangalore to join me as a co-host, hi Krishna.
Krishna: Hey Zoe, thank you so much for inviting me to be a co-host on this episode.
Zoe: Yes, so this episode, we have the pleasure and honor to invite Romain who join us. And then I would like him to addressHow are you? Let yourself to introduce, okay.
Romain: Yeah, so my name is Romain Bouqueau. I've been mostly known for being an open source contributor on the GPAC open source project. So that's a multimedia project, mostly known for packaging streaming distribution around the ISO BMFF MP4 file formats. So I think we're going to say more about this. So now, I start to be a veteran in this industry, I've started 20 years ago in several companies. Before, 10 years ago, I started my own business around the GPAC open source project. So providing professional services and licenses.
Zoe: So yeah, we are talking about GPAC. So this is something you have been dedicated to. And then recently, not recent, so GPAC, and then together with the, I think, Motion Spell that you are leading. The effort there. And we like to learn more. I know that most of the audience could know GPAC well. And then especially for us, like we use MP4 box almost on a daily basis. So you may want to still put a little bit of effort and time, let's discuss about this, have some short intro about all these open source platform softwares that actually influence our daily life, yeah.
Romain: So it's really interesting in the multimedia area, because we have many open source software that were created in France exactly at the same time. So we have GPAC, we have FFmpeg, we have VLC. We have MediaInfo. There are many other French project, like TSDoc for example, for manipulating impact TS. So there was funny things that happening at the time. And for GPAC, this was made by a trainee actually, who left France to go in New York City to make a startup. And that was the .Net bubble. It's just, so the bubble exploded and the guy came back to work in a university. So this person is Jean Le Feuvre, he's the creator of GPAC, and so the open source, the code that had been made for this startup. And that's the beginning of the story of GPAC.
And what is funny is that at this moment, there was a merge of several audiences. What people have to understand is that 25 or 30 years ago, people had visions of what computing would look like. And it was like second life, some VR environments with a lot of interactivity, immersiveness, real time. So they had the vision of what the imagining could be. And so we had this new technology, like Macromedia Flash, and MPEG at the time said we need a set of standards that would be open standards as an alternative that would be better, and so they made MPEG-4. So MPEG-4 is a set of standards. There are some video codecs in it, and the most successful is H.264 AVC, MPEG AVC, that's ubiquitous.
For audio, there is AAC. So it's a set V MP4 file format, of course. But that's also scenes, interactive scenes with some scripting, with some text, subtype, you know, streaming of text. And so many other things like metadata that, you know, are still used or not up to today. And so, when GPAC was done, it was not only 2D video, like TV-like experiments. And what I was going to say, that there was a merge. Like, the users of the failed startup followed what GPAC was doing because the people were interesting, the project was interesting. So some people started to use it when it was open source. The creator of GPAC went to work in a university.
So they were researchers, they were students. And that was the big era I think of, you know, piracy. And so there were a lot of people, you know, starting to grab the videos and putting them in a new MP4 container that looks so promising. And what is funny for us is that GPAC has evolved. When I joined it like in 2007, we were doing still some 3D crazy stuff. And eventually, we discovered that some of the pirates became CTO of big companies like YouTube. And so they came back to see us and they said, "Guys, would you do some business with us because we just run into production and now we have production problems?"
And I think that's part of the things that triggered the creation of, you know, GPAC licensing and the Motion Spell company at first. It's just we had, we could cover everything that was research and R and D, but the maturity and the level of service and the speed at, you know, closing the deals, et cetera, was just insufficient in the university. If the university had been good, I would be probably a researcher and teaching in the university.
Zoe: Yeah, well, different sectors have different expertise. I think the researchers mostly always look into something really deep and then predict what's going to happen maybe longer into the future.
[00:06:29 The Strength of GPAC]
Romain: Yeah, so that's one of the strength actually of GPAC. I think, in term of R and D, we are so much ahead of many of our competing projects. And so, it's really interesting I think for companies because most of the time when they need a new feature, it's already in GPAC. For example, when Apple released their headset, they decided to use the multi-view MV-HEVC codec, and it's been in GPAC for years. So you just had to take a video from the headset, you could edit it, you could edit your own videos. It just worked, you could distribute it, you could dash it, you could encrypt it.
So I think R and D is a big differentiator. Also another differentiator I think, compared to the other open source projects, I think open source is trendy. There is a lot of companies doing open source, but the problem is if you don't have a community, like, you know, real people dedicating in their free time, using it, maintaining it, it just doesn't work. Like, companies have created open source software. And at some point, when nobody's willing to maintain it, the open source projects just disappear.
So if you look, for example, there are two other, so we do mostly packaging I think as a business. And there is Bento4 and Shaka Packager also operating in the space, but the problem is they've been created by companies or individuals and they have no maintainers anymore. And so the question is, what do you do with it? It's always the problem of sustainability in open source. I think GPAC has a really good balance because we have, you know, all these massive R and D, we have this massive community. And so, it's my role to make sure that, you know, everybody's happy, but that makes the project really sustainable and innovative and flexible.
Krishna: So just picking off that last sentence of yours, Romain, community, maintainers, and contributors. How do you get these three right to keep open source things sustainable?
Romain: So we have an arrangement. Basically when I created Motion Spell, what I wanted to do was to attract the GPAC talents so that they can make it a living. You can contribute to open source and get some money back, that was the idea. So we managed to do this up to a certain point. I think it's really difficult to retain talent, I think talents is a problem for any company, and that was the case for us. And we used to be like a full remote company, so attracting a lot of people. But then with COVID, a lot of companies became attractive as well because they were remote. So that was not so much about the differentiator.
And about redistributing the gains, it really works well. That's really a nice way I think for people contributing to get some money back, that's also recognition. You see your name on stuff, so if you're looking for a job, you see your name attached to a specific feature on a project that is as big as GPAC. So, but yeah, for me sustainability is something that is really, really important. And I think over the years, people have become sensible to the stories where they all rely on an open software that's been maintained for 20 years by one guy somewhere in the world. Not sustainable at all, but we don't have this problem in GPAC. And 10 years ago, I created this company just to make sure that it would be sustainable for the people willing to make it a living.
Zoe: So this way, actually that's what we enjoy, the existing functionality and features, just the two parts. You always see something that maybe that you need of, and which has not been on offer yet. So that's basically, we want some new things added to the existing. On the other side, even for the existing feature, there's always a different way to use it. And then sometimes, especially when you actually work on a lot of different use cases, there may be some course case that has not been covered. That's always the case. That's why even the most advanced companies, they have the patches and patches into the existing release. So this is some support that also needed by the open source software. And so you mentioned sustainability, it's really, really important. The community really rely on this and then they also return, help build the confidence for that.
[00:11:33 The Customer Journey]
Romain: Yeah, yeah, that's true. So about the contribution, I think, some people, some technical people are really open to contribute. But when you have to deal with businesses, it's a bit more complex. You are an engineer inside a corporation, you want to contribute, you depend on the lawyers of your corporation. So that's where we try to make it as easy as possible. So there are contributor license agreements, but otherwise we just propose professional services.
And in most of the cases, just so much easier for companies to pay for someone else to do the stuff and not have to deal with two things. First, the legal part, but also the maintenance. 'Cause when you pay for development in GPAC, it's going to be maintained forever for free. If you do it internally, you have to maintain it, your teams have to maintain it, and this has a cost. And then there is the second point that you raised, Zoe, that is super important, that's the usability in general, like the use cases, et cetera.
So we are trying to address this in different ways. We're trying to continuously get feedback, that's really difficult because open source software don't have analytics. Basically GPAC and MP4 box, when you download them, they are just tools that you use, and we don't get any analytics from you. So we don't know how you use the software. And as a business, that's a problem because you want to know, as much as you can, what is the behavior of your users and customers?
Yeah, but we have all the ways to do it. We have a website where, you know, we have cookies and we try to try the user journey. And it's important for us to have good documentation, and more professionally to propose trainings. So you're a company, you want to use GPAC and you can buy trainings, and we can propose something that is more custom and personalized so that, you know, you step up. There is also something else that we would like to do and, you know, we try to go this way.
I think there is a lot of, and it's similar for any technical tool, I think there is a lot of friction. It's really difficult, when you have a new tool, to go from zero, I need this tool, I know I need a packager, or I need to stream, I need to encrypt, I need to put advertisement, up to the point where you are operational and you are independent in the way you operate. And so, okay, we can train you. But then if you need us again, because I don't know, you didn't understand something or whatever, you have the documentation, but you need to pay again.
And so, of course, there is the community, but there are so many problems in big organizations, like the teams change, et cetera. And so, I would like to mitigate that. So we've started to lower the friction using web assembly. So web assembly is a way to run some codes directly in the browser. A project like GPAC has two types of, you know, community users. They are developers and we have power users, people not able to develop using the command line. That's the same for FFmpeg I guess, and many other tools. And the problem is, you need to go to the website, you need to be able to install the tool. If you're on a platform that is not supported, you need to build it yourself, it just creates so much friction for so many users. Web assembly allows us to run in a browser.
So, of course, it's not feature complete, but it's great because when you're in a browser, you can also present it in a different way. So we have the wasm.gpac.io, you can try it for free. And you have some presets that allows you to launch, directly, GPAC from your browser. So I think it lowers the friction for people to start using it. And I would like to go a step further. I think that a lot of these tools, a lot of packagers, or even FFmpeg, there are not so many people using it every day.
At Visionular, you need it like every day. But in some teams and companies, people use it, they do their stuff, and then they forget about it for quite some time. I think we just miss a place that would be even more user friendly, where people would get the information about GPAC. But it's not so much about GPAC, it's about solving your video problem. So that would be a graphical place, probably in the web, where you solve your problems and you have a graphical way, like with a user interface, that makes it more, you know, that leverage the social part, the community part of GPAC. And that helps you, you know, describe your problem and find solution to your problem.
And, of course, yeah, operating this as a company means that, you know, it's a kind of community place that just leverages the open source software, or I should say several open source software, because GPAC leverages other open source software as well. And, you know, putting the focus on solving the problems of the users, but making it so much easier to use. Making people come back to this place again to solve their problem so that, you know, it's not like, "Oh, two years ago I solved the same problem, but I don't remember how I did it." What was the name of this, you know, project?
Krishna: Yeah. With the advent I think of, with the ease of large language models, I think you might have a pretty easier path towards this if you can feed it all the documentation and actually train it to give answers.
Romain: Yeah, absolutely. So that's part of the solution. 'Cause most of the time, the journey of the users is, I have a problem and I ask for some answers. It can be to, yeah, an LLM, it can be ChatGPT, can be Google, can be, you know, whatever place. Or their colleagues, or... And that's the moment where, from the GPAC perspective, of course, I'd like to have more users on GPAC. So I'd like this entity to answer, have a look at GPAC.
But the problem is, that's an SEO problem. So that's the first problem that we need to solve. You know, how can we be visible? So we can improve this by, as you mentioned rightfully, train an LLM on the documentation, just to make sure that at least the people reaching GPAC can find solution to their problems. Because for us, that's the second step. People discover GPAC, and then can GPAC solve their problem? Yes or no? And that should be clear. 'Cause GPAC is really abstract for some people, it's just packaging, you know? That's nothing visual. So you arrive on the website, you're not sure it's the right tool. We wanna make sure that people, you know, easily know if that's an answer to their problem, yes or no? And if it's yes, I wanna make sure that people remember about us. That's the customer journey.
Zoe: So back to, again, we are here talking all about GPAC as well as all the services and support behind Motion Spell actually to, I would say, not only advocate GPAC, but mainly you mentioned the sustainability of a open source software, and that needs actually by the whole open source community. So you basically really facilitating this vision and mission for this open source. And this is actually, we got some questions, like they say, "Oh, this is open source, it's supposed to be free, but then there's a commercial company." And then they usually will think that the other ways is a commercial company is leveraging what has been achieved by open source to support them.
But here we actually realize together how this commercialization effort to actually in return to support, to actually make a long life of this open source community and software platform. So can you also give us like a, I would say, a big picture of Motion Spell? And give us audience, what exactly services and products that it offer? And then how is actually support? You just mentioned several points already, the developers, the users, and including individual users, and as well as the big player, like big ones. They also want to, let's say, YouTube came to you to want to leverage this. So can you give us a big picture, and how Motion Spell is actually providing?
[00:21:02 Big Picture of Motion Spell]
Romain: Yeah, so that's an entry point for companies. I think lots of companies, open source, I've seen the change over the last decade. 10 years ago it was like, "Yeah, no, I'm not going to use open source." And these companies just became dinosaurs, because the technology has evolved so fast. But open source commodisizes everything, right? When you have a good open source option, it's just going to take a large share of the market. Look at FFmpeg, when it comes to encoding, in any team, there are people using FFmpeg. If you need to package and, I don't know, you need to encrypt your content and you need to package it for Dash, you're going to use MP4 box. And I think that's really one of the strength.
Now the problem is for some companies it's just impossible, or they have larger needs. So they want to have some support, they want to have some trainings, they want to have some custom developments. Companies have many questions, and they don't operate at all like a group of students a group of passionate people. So you need to find a right balance as a community on, who can give what to the community to make it sustainable? For me that's the key word, that's the key word. And to what we offer, so whether you're a small or a big company, you can contact us, we have a contact form. Some people contact us on LinkedIn as well. And most of the time we just operate as any service company. So we try to understand what is the business need, and we try to address it the best way. All the rest is transparent for companies. Companies don't care if the people entering are employees, freelancers, people from the open source communities or whatever, we just make it transparent. It needs to be easy for everybody.
I don't know if it answered your question. But, for example, for Netflix, they contacted us, they had questions. And at some point they said, "We want to continue working with your open source packager." "Is there a way to have some professional services on the top of it, so that we start deploying and we see how it goes?" And yeah, hopefully it went until full deployment. So that was really great for us, and we got this press release that's really a highlight for us, because companies using open source, most of the time, they just stay silent about it. And Netflix were so nice to communicate about it, so that's good publicity for us.
And I think it opened doors. So for some companies doing, you know, shamefully some open source stuff and not saying that, you know, it's not shameful. For a company like Netflix, they just said it openly, they're not competing on this. They're competing on other stuff, like the content they have, and how they serve their users. It's not necessarily on their packaging.
Zoe: Yeah, especially you don't have to remake the wheels. I think that's the actual leverage-What example GPAC and MP4 box is provide.
Romain: Yeah, again, I think it's a combination. Yeah, if you do it in house, you have to pay for the development, you have to pay for the maintenance, you have to do the R and D. And that's where the thing that I said before, if you have an open source software, that everything becomes a commodity. And I think as I said, we've created the right balance. We have the R and D, we have the right structures, we are not too big to have problems, but we are not too small, so that we are really sustainable with enough contributors.
So I think we're at the sweet spot that makes it super interesting to work with us. That's my opinion, and that's what I see. Companies actually, we have moments where there are too many companies contacting us at the same time so that we can serve them, you know, well. So we're trying to find solutions for companies, like sometimes they don't need professional services, but they need more training, et cetera. And with the initiative that I mentioned before, like, you know, creating something that would be more graphical, et cetera. We're just trying to find the right balance between the talents, the knowledge, and the availability that we have and the actual needs that companies have. It needs to be affordable. We need to have a solution for any kind of, so on the Motion Spell side, for any kind of customers that would come in to see us.
Krishna: And what you mentioned about Netflix is also a good example of where a company paid for an open source, and supported an open source project. I guess the other option is for companies to submit bug fixes. What is your perspective on this, Romain? It's an open secret, everyone uses open source, goes to GitHub, finds something, downloads, it uses it. What is your suggestion on how corporations can give back?
Romain: Yeah, so I think the best feedback is actually the user feedback that we don't have because we don't track the users. So that's the first thing that we are looking after. So I think we have a better view of what the professional users do, because they use technology in a different manner than the other ones, like, especially in the streaming media industry. So in the streaming media industry, people are talking about personalized advertisements, DRMs, these are things that don't exist with our other customers. Like, if you use GPAC, for example, for display in transportation, in trains, in hospitality, in elevators, for medical equipment, you just don't care about DRMs, you just don't care about advertisements. If you are someone just grabbing the videos from your camera and trying to stream it for, you know, small events, if you're working for a church or whatever, you don't care about this.
So there are real easy ways for us to segment parts of the users. And then the professional users, they don't react the same ways as the other. They don't behave the same, opening GitHub issues for communities. When they send emails they don't, you know, behave the same. There are some slacks, there are some, you know, spaces that are definitely open to professional users. LinkedIn, for example, I never received, like, some open source contributions on LinkedIn, doesn't exist. But some people send comments or they send messages on LinkedIn. So it allows us to say, hey, there is some interest. For example, there's been a lot of interest in just in time packaging and transcoding recently.
So I think that's because of the, everything that is green. People just realized that they had appliances running 24/7, they just didn't need it. They just didn't need to transcode all of the good qualities and package for all the devices, that was just useless. And so they said, how can I just generate on the fly when people actually need it? Like, there are tons of people generating for the PS3, the Sony PlayStation 3. You know, how many simultaneous viewers do you have on this device? You know, or just for old Apple devices, just supporting a VC baseline profile. You generate all the time. But if you consider like that you're a content provider, for example, you're providing live or, you know, linear and you have 200 channels. The least popular of your channels, you don't need to transcode them in all the formats all the time.
So there is some popularity on this. And again, this is a use case that is typically for people trying to make money out of the video, that's the streaming media industry. And so, you know, we can identify them and we can best serve them. So I'm trying to make sure that the professional users, their contribution is mostly money. That can be pull request also. But as I mentioned before, there is so much complexity legally for companies to have engineers contribute. But I have another customer, unfortunately I cannot name it, but probably we're going to make a press release before the end of the year, where we're receiving pull requests all the time. They send us files, they send us detailed reports on problems that's so valuable for us. We're just trying to create the best software. We're just trying to understand what people, you know, use GPAC for.
And so at some point, it's not only money, it's just building the best possible software. Because when you have the best possible software, then it's just a black hole effect where everybody, you know, comes to your project. But again, it's not only technical, it's also having the documentation, providing the right level of, you know, support to all the users reaching you. And again, packaging is a niche. So I'm not trying to battle for SEO, you know? It's only the people typing for the problems that GPAC can resolve that are going to come to GPAC.
Zoe: Yeah, you basically mentioned at this one point, for example, for the open source, we all know that there is open source community, these lovely users, they will actually give their and share their feedback. But if you have commercial users, like the one that you just mentioned, and then when they are willing to pay, they basically bring up the two extra market request. And then that request got back to you and then you'll learn, and that actually add the true request for the features and functionalities that are needed by the market. And then this will actually help iterate the product and offer from your side. So it is this kind of closed loop mechanism that help iterate the offering, as you mentioned, that really helps sustain the community because it's actually moving forward, and also address the real needs in the market.
[00:31:48 User's Community]
Romain: Yeah, but to do that, we need the awareness of the people that this is open source and this is a community effort, 'cause it's free, right? And I'm on the worst possible business, I'm trying to sell something that is free. So I just wanna make sure that people, you know, people appreciate us because we're approachable, we are knowledgeable. And they know that if they have impossible problems, they can reach to us, right? But it's not only this for me, if they needed training, they need to think about us. If they need some support on some stuff because they're using GPAC and they have problem, they need to think about us.
And this is because as a whole, as a community, because the community is not only the people doing GPAC, it's not the core of GPAC, it's all the users. If we behave responsibly, then it's going to be sustainable, otherwise it's not sustainable. Like, we're not a company, you're not paying a high price. Most of the users are just free riders. Really recently, I was trying to get contact, some contact in a specific team at a really, really big streaming companies, maybe the biggest competitors of Netflix. They came on our public issue tracker and they were tracking the problems, not providing any sample, and they were behaving a bit aggressively.
So I was trying to send messages to people on LinkedIn and saying, "Hey guys, you know, maybe, I don't know, we can do things together, like, you know?" "Maybe you're not allowed to share your content, and that's the reason why you don't share it or whatever." And they just remained silent. And, you know, what can I do about this? You know, most of the commercial companies don't have the problem. If you don't pay, you don't use the service. But GPAC is an open source project. So, okay, we have to deal with it. And that's the same with the other communities.
It's not because that people are passionate, that they are going to be, you know, better contributors than the industry. You know, it's just separate, you know, communities. It's invisible for most of the people. When you work in the streaming media industry, you are always dealing with the same, you know, one to 10,000 people. And basically I would like to say, we know each other and we've known each other for years, and, you know, they are newcomers, et cetera. But we've been around for years dealing with this technology, but GPAC is more than that. You know, I have customers, a couple of years ago, the streaming media industry was less than 50% of my business in Motion Spell. So GPAC is definitely not only streaming media, because the streaming media industry has difficulties to innovate.
At the moment, it's a kind of shrinking market. Most of the R and D is elsewhere. I was discussing with Fabrice Lorenzo, the ex CTO of Lifelike Boat by Cosim, because they, you know, they have this spherical display. You know, giant sphere, where, you know, there is one in Las Vegas but they have, you know, other ones in many big cities in the US where you go to see your live sports events. And I saw people complain where we're talking about, you know, that's greenwashing stuff, et cetera. That's the same for Formula 1. For me, these use cases are purely innovative use cases of the things that you may see in the future in the streaming area. For me, that's a super interesting experience. So, of course, it uses a lot of energy, right? But if we hadn't this, you know, who would do the innovation in the industry? Think of most of the big companies, most of them have always zero R and D, into push innovations. In front of big companies just, you know, rolling marketing plans, you know? They want you or they don't want you and, you know, you just depend on them.
As small innovative companies, yeah, for example, for Netflix, I think that people ask all the time, why did they choose you? And that's the case for all the other companies I can talk about and the one I can't talk about, that's the flexibility, you know? You have this open source brick, you can do whatever you want, you can extend it on your own, et cetera. As small companies, we're really flexible. That's what we offer also to our customers. We can run on their own appliances, we can run in the cloud, we can run on a specific processor. We can do things that the big companies can't do.
Krishna: That's amazing.
Zoe: Yeah, thanks so much Information part, yeah.
Krishna: Yeah. Romain, so now that you mentioned Netflix, could you take a couple of minutes to talk to us about that engagement? Because I think it's a huge win for GPAC. How did this entire engagement take place? What are you actually providing? How does GPAC sit in the Netflix infrastructure? Can you help us understand a little more of that?
[00:37:00 GPAC and Netflix]
Romain: Yeah, so they are our sponsors. They give some money just to make sure that we're sustainable. I think any company using GPAC, or any open source project, should consider having a fund where they give money, even if not a lot of money. Some years they may not be able to give, and some years they may be able to give. But I think it's important to give back, and for companies giving back money when it's possible is an option. And then they buy some professional services. They need to investigate all the time in the R and D.
And Netflix is not that big company. I don't know what is the right number, and maybe you can have a look at that, but the technical teams are not that big. So actually they rely on us for some of the stuff. Can you investigate, you know, how it would be if we had to use DXC's technology, or if you wanted to do this or this? And so, they buy professional services. And most of the time, it's, you know, it's our passion and Netflix is definitely an innovation driven company. So that's, yeah, possibly one of the best customers you can dream to have. But there are so many interesting use cases, of course, as well. I don't know if that answers your question, or if you have additional questions.
Krishna: So what exactly does GPAC do in the Netflix framework? How did your entire engagement start? You can just walk us through that journey.
Romain: Yeah, so I gave a talk on stage at the NAB Streaming Summit by Dan Rayburn last year and there was an engineer from Netflix on stage with me. They use MP4 box for everything. They use it for processing the production files, which are stored in ProRes. So ProRes is basically a derivative profile of the QuickTime file format. And basically ISO BMFF, which is the basis of MP4 and CMAF, is a really close clues into this. So they use it for production, they use it for distribution to the end users, and they use it also for the journalist previews where they have specific constraints. And it's basically all the use cases that they have for video, so we are everywhere.
Krishna: That's great.
Zoe: So right now, GPAC, not only just the several sectors it's actually serving. So we see broadcasting and see, like you just mentioned, Netflix and in the OTT streaming. And then for a lot of users, there is definitely for different purposes like user generated content. So serving a lot of things, we do notice that for your effort, 'cause this is closely related to new standards. So you also mentioned the dash, and then fragment data like MP4. And then we like to actually learn how you actually got engaged or how the relationship between your effort and the standardization, standardization community? Because they actually tried all to absorb the new things into the standards and moving things forward. But then for you guys, you need to actually support all kinds of different standards, but there's also emphasis in the market needs. So we'd like to learn a little bit more how Motion Spell and GPAC relate to the standardization of this streamings, as well as even inside like a codec, HEVC.
[00:40:52 How Motion Spell and GPAC relate to the standardization of this streamings]
Romain: Okay, so we've been long time standard contributions, contributors. We have more than 500 contributions at MPEG, I think mostly in the file formats group. That's because GPAC has roots in research actually. That's what I said, it's something that's missing in a lot of companies and a lot of open source software, where they don't know which technology is going to arise just because they don't make the effort of following it. And all companies, like a couple of years ago, companies have started to send people to MPEG and to W3C and to conferences because there are many conferences that are, you know, more industry-oriented, such as Mile High Video for example. That's a conference with real research papers. And so the other conferences is starting to do this, like IBC.
But for us it's, you know, it's not, it's not black and white, it's not you are doing R and D, or you're doing standardization, or you're doing this or this. There's a continuum of actions that, you know, makes your, you know, how do you say it? Actions consistent, you're building stuff from the ground. But when it comes to, so we are pushing the standards and we are following some of the stuff, we cannot follow everything, and in GPAC we cannot implement everything.
So like any project, we did some good bets and we did some bad bets. But then on the Motion Spell side, on the commercial side, we also have trusted customers that are big consortiums, like 5G MAC for XR, or CTA Way for representing end users in the US. Or we have a smart city collaborative research project. Or we are going to be at the IBC Accelerator this year, about ultra latency on a project that is championed by Comcast. So that's where we're trying to do the link between the things that we understand, and in research and implementation.
And when you think about GPAC as an innovative project, what is great is that it's really mature. It's a research project that is really mature, and that allows you to go to market faster. So we have lots of customers actually who say, "Oh, I would like to go to VR, but I'm not sure it's going to take off." I start with GPAC, and I see if I have customers. It's a pure, you know, lean approach where you put a proof of concept, you use open source to do it, and then you see what happens. And so I have many customers having licenses and professional services, or taking agreements or taking trainings just to be up to date with some features. That's a missing link, right?
If you do only deployments, you're always late, you know? For example, recently we got many discussions with AOM that were related to AV2 and the conformance of AV2. If you're deploying AV1 now, you're late to the game. AV1's been done, what? Six years ago, it was implemented in GPAC. So if you're doing AV1 now, you're six years late compared to the actual work. And that's what I said, for the GPAC users, it's great, you know? You're doing AV1 now, AV1 is mature in GPAC. You can take it, and you can push it in production. YouTube's doing AV1, Netflix doing AV1. It's mature, it's deployable right now.
So I think it's really appreciable for the customers because most of the time, they just don't need to pay for an additional development, it's already here. And when they need to pay for an additional deployment, that's because, okay, you've implemented AV1 but I'm using AV1 with this specific metadata with this new HDR technology or whatever. You just, you know, put your brick on the top of what's existing, and then you can deploy on your own. We try to make it easy and fluid and flexible.
Zoe: Yeah, so you mentioned not only GPAC is a enabler of lot of product and services, it's also the test bed because it's rooted. Not just the innovation, because this is rooted on R and D from the very beginning. Because you also mentioned MVHEVC, right? Previously, in other conversations. So MVHEVC is a there for a while, and that is also inside MP4 box. 'Cause right now, we also mentioned when people mention, "Oh, we want to do a video that can be a special video that can be played back on Vision Pro." And with kind of tools, I believe people think the first one is MP4 box on top of GPAC. And so, that's actually addressing the VR sector. But you also mentioned, you can't choose everything. So we are wondering how this one, MVHEVC, right now at least empowering the potential of VR/XR in the special video world? And this GPAC has actually played a critical role down there, and you chose it in the first place that has been there for a while.
Romain: Yeah, so I don't think there is any future for MVHEVC, to be clear. It is just encoding one eye, and making some prediction on the other eye. So people are now going, actually you can use MVHEVC for multi-angle cameras, et cetera, but you have other options. It's a bit awkward, right? As you know, I remember in 2008, I was already in this industry and we were, you know, streaming on 3G, using RTP and AVC. And then Apple arrived and said, we're going to do HTTP live streaming using Impact TS.
And we were like, "Dudes, this is all wrong." We're just going to waste our bandwidth, you know, what is this? And it took us how many years to deploy CMAF instead? And now, we we're only talking about switching to media over quick and quick and HTTP 3 that wouldn't rely on TCP and, you know? 15 years in the middle.
Yeah, you know, you need to do it. So let's make it work and let's make it correctly, but then we just wasted 15 years, that's a choice.
Zoe: Yeah, thanks for providing at least one perspective of this.
Romain: Yeah, but that's because, when we did it, actually we were part of the standardization on this, and that's why that we implemented it almost 10 years ago. There wasn't the traction, the traction was on 3D. And 3D never took off basically, because the people don't want to buy an expensive headset and just put it on their face. The best use case for it is not video, that's video streaming, that's video games.
So as I said there are other use cases for video, but outside the streaming media industry, that's not a mass market. VR is popular in the industry in general, or in retail, or you can do some training, you can immerse people in, you know, specific situation. We have lots of customers on the medical environment. It makes sense, but it doesn't make sense for me in the streaming media industry. That's my take.
Krishna: At the current price point, it's a little tough for sure.
Romain: Yeah, you don't need to use this device, and probably the other devices may be compatible with the content that is generated for the Apple Vision probe for sure. But do people want to use this technology? I'm unsure, they've been using 3D, using other technologies as well. Not sure.
Zoe: Yeah, so basically you at least mentioned that, for example, in the medical imaging world, where right now like a lot of learning environment, so people will think okay, if you have the 3D demo, that's more intuitive for the students to not just learn, but they feel and they could see. And so, different ways that present the knowledge.
Romain: Yeah, so yeah, absolutely. That's, I think that our industry focuses too much on the acronyms, et cetera. And we're creating virtual barriers. People are just interested in solving problems. They have problems and issues, and they need just to solve it. And in a medical environments, they just don't ask if that's AV1 or a motion GPAC. Most of the time they just don't care, just wanna make sure it's low latency. They know that sometimes they send, you know, difficult images for codecs. So, for example, they're interested in HDR, but not for the same reasons as the reasons of the movie makers. They're just interested in this because if you're looking at an image, you wanna make sure that you can differentiate some areas from some others in surgery, for example.
So these people, when they come to see us, they don't say, "Hey, I want impact dash with this and this." They just say, "I need to stream images, and these are the sample images that I have, and I have a constraints that it must fit in this number of megabit per second with this latency." "What can we do?" That's all they say. It's not like, "Hey, I need to avoid this codec because there are patents, and I need to avoid this because it's not going to be hardware decoded by this device or it's not deployed on AWS," or, you know, whatever. They don't have this level of constraints, they're just concentrated on the problems. And I just feel that our industry should do the same. For example, they're not able to deploy any innovation, let's, you know, make it clear.
But when it comes to making money, they can deploy anything. Like personalized ads, we can do it. And they make something that is not convergent, let's say, at the moment it's a bit complex still to deploy. They are converging, but super slowly. But at the moment, you have ad insertion and you're tracking the user, et cetera. Why don't you use this tracking technology to provide something that is super useful for the users? Why don't we have like in-app or in-video purchase for any use case? Why don't they put the focus on this? Why don't they put the focus on not only providing advertisement, but also information? Like, someone is talking about something and you see a link where you can click and you can get additional information. Why don't they provide this? They all provide customized advertisement that they're not able to provide this to the user. That would be, you know, added value to this.
When you are watching your content, it's playing 2D video. It's like your TV 40 years ago, that's the same. It hasn't evolved, the user experience hasn't evolved. For me that's the real problem, they're complaining that the users go away and they play video games, and they prefer browsing the web and they prefer... Yeah, right, just improve the user experience. If you engage people, they say, "Yeah, low latency, people are not asking for this."
Well, I don't know. If you put low latency and people are really engaged and they can ask, you know, questions, et cetera, if you can make it engaging, people will follow. For me that's just, you know, a trivial truth. It's just, it's not done correctly by the actors. They're not, you know, putting the money in this. They're just putting the money in different areas, like creating always the same 2D video content and then complaining that, you know, nothing happens.
But they are really innovative companies. We have, I'm going to tell you, we have customers providing, engaging their audience for sports. So you're watching sports and they're providing, as an overlay, you can hoover and get some statistics on the players and system replay on similar games, et cetera. That's values, right? But we need that for all the contents, not only for sports.
Krishna: Right, in at least for the IPL in India. MyIPL in India, they had multi-view, you could switch the camera angles, a 360 degree camera on top of the stadium. So you could see the match from above, pretty decent innovations I would say.
Romain: Yeah, absolutely, absolutely. There are tons of use cases that are interesting, but, yeah, it's... I don't know how to say that, but for me the not traditional, but the people in the streaming media industry don't realize how much the lack of innovation is a problem for them. They have the feeling they're not competing in a fair way with the companies from the Silicon Valley, but it's not what's happening. Traditional broadcasters, they have tens of thousands of content in their catalog that they can use for free.
But they need to do the work that is necessary to refresh it, to clean it, to present it to the users in a really compelling way so that people, you know, come back and see their contents, you know? They could make an alliances to altogether to propose international content, et cetera. It's not like, oh, Netflix is creating content and they are closer to the users, they're producing all over the world now. Just you've been producing all over the world for 30 years. You know, don't tell me you cannot do that. You see the other broadcasters, you make an agreement with all the national broadcasters from the other countries and you have a catalog of, you know, hundreds of thousands of content that Netflix will never have. You know, what is the difference?
Of course, you need new content, you know, and, you know, better content. And we don't do content, we don't create content the same way as we did 30 years ago for sure. But people like classics too, you know? They buy companies based on advertisements, having a not new catalog only, you know, they rebuy old catalogs. Like Pluto TV and, you know, this works. Or when you go to Disney+, for example, because you have kids and you want to see the old classics movies.
So for me, I'm not a specialist in all this market stuff, but for me it's doing innovation. I just see my biggest market is streaming media, and my biggest innovation market is somewhere else. There's a discrepancy.
Krishna: Yeah, that's a good insight, yeah.
Zoe: Yeah, basically we can tell that you are the evangelist of open source innovations. And that on top of this, there is usability and user experiences and addressing the real world issues down there. And then, so we like to actually learn a little bit of you, because right now everybody talking about AI, and then there's a lot of things going on there. Like we already touched the model, and then that will partially... 'Cause we know that some of the standardization activities we're thinking about that. And then, so you are running this community to support that. There's a lot of potentially gen AI videos that will also come into being, a lot of weight, that it influence every single part of the industry, as well as our daily life. So what's your view related to streaming, and the things that you are engaged at this moment? Yes.
[00:58:00 The influence by Gen AI to Streaming]
Romain: So, yeah, so, first it's not my job. The more video we have, the better for me. AI generated, user generated movies, wouldn't make a difference. That being said, when I saw ChatGPT for the first time, so my wife is a graphist. And so she's been following the mid journey stuff, you know, evolving from version one to version two to version three, where it just got, you know, kind of uncanny to most people because they just said, "Hey, on the creativity stuff."
For most people, it does the job instead of me. So it's really funny because AI just replaces that for the creative stuff. And so, I was really scared and I said, "Oh, should I go into this?" And at this moment there was no CPU implementation that would be open source, and so I dug into it. I remember that was at Christmas holidays, I was at my parents' place and I starting to develop like crazy trying to understand how the inference would work, et cetera, and doing the training.
And eventually some other people took over and I came back, just because my only conclusion is that we, I think we overestimate the human intelligence. That just says that about us. We just says, "Whoa, that's magic." But again, when you think about it, people are replaced by machine, not when the machine are perfect, but just when the machines are better than humans. And this is just the inflection point, you know? You can replace human drivers by robots as soon as you can show that the robots are better than humans, that's all.
I think that's all people care about. You can replace the driver of a plane as soon as the driver is better. You can replace a human on the robot as soon as the robot is better than the human. For me, when you're thinking about this as a thought experiment, that's when it's happening. And this is what has happened with the DLLM in term of language. They just talk better, they just write and read better than most of us. You know, I'm not a native English writer. It makes a perfect sense for me to ask ChatGPT or a competitor to write some English stuff for me. It's too complex for me, that's a foreign language. Even though I talk, I'm talking English all day, it's so much easier, they are so much better than me.
So, of course, I need to check for the semantics because most of the time it's just, you know, creating stuff or putting a false fact on it. But for me, it's not the problem. And for video, that's the same. As soon as the video looks realistic, there's going to be some new video that's going to be created. Up to a point where, you know, there's a problem. Like the web was user generated, we trained on the content that was user generated, and suddenly everything is flooded by machine generated content. So the problem is, what is going to happen now? 'Cause the new models are going to be trained on, you know, machine generated content. So what's going to happen? Yeah, but for me, the natural way is this, but we are far from humans consuming machine generated content, like machine generated video.
I think you would never pay for this nowadays, and that's going to be the same for quite some amount of time. But if, for example, a small company like mine or Visionular wanted to create an advertisement that is a video, I'm pretty sure we could use it to make it cheaper to do it. Definitely writing blog articles, that's what I said. We are flooded by fake content at the moment. I think it's impossible right now to read, there is no genuine content anymore. Most of the companies don't write their own content. They just say, I have these ideas and say to ChatGPT, "Provide some content that is a blog article." And then I say, wow, that's too long. So I ask ChatGPT to say, what are the main ideas of this article? This is becoming stupid, that's my take. So I don't know, but for me, it's just more videos to process. So it's good for business I guess.
Zoe: Well, thank you for actually sharing this. And now we actually, even myself today, learn a lot. I think the thoughts, motivations, and then the mission and vision behind GPAC, as well as the Motion Spell business that you are running at this moment now for quite a long time ago. And then for the keep the sustaining of GPAC and we know that all this effort will ongoing. But then as you mentioned, new content will come, and new requests emerging from the market, and innovations will always be there and then push everything forward.
So I'm going to wrap up. We had lot of discussion today, but based on what we discussed today, we'd love actually down the road to invite you back to our podcast to share what we discussed today, how they evolved. And then, 'cause we look at the streaming industry, we look at a lot of new video content, and then you also mentioned even you think AV1, who actually took us still just to take off, but it's already six years old. So new things are coming. So we'll see what's going to happen in the next near future, and then we can actually come back and discuss this topic again. And I really, really appreciate all your sharing with us today.
Romain: Thank you, Zoe. Thank you, Krishna.
Krishna: Thanks also-
Zoe: We actually thank everybody to be with us for this episode. Again, thanks a lot for us to learn about GPAC, learn about Motion Spell, and learn about all this open source, as well as the effort behind it to, again, innovations is a key and that's actually drive everything moving forward. And we're going to enjoy the even better world, definitely down the road. Thanks everyone to listening our episode, we'll see you next time.
