The VideoVerse
The VideoVerse
TVV EP 27 - Flavio Ribeiro: The Past, Present & Future of Video Streaming
Tune in to hear Flavio Ribeiro, Sr. Engineering Manager of Netflix’s Live Streaming Technologies, discuss all things video streaming. Starting in the streets of Campina Grande, Flavio shares his journey from contributing to the recreation of Brazil’s digital television system and working on Globo’s live streaming platform for the 2014 FIFA World Cup, to helping pioneer video processing and delivery in “The Big Apple” at companies like The New York Times, Paramount, and now Netflix. Not only does Flavio share his experience and the things he’s learned along the way, but also his thoughts on what the future holds for video streaming. What will be built today that will continue to be used a decade from now?
Welcome to "The VideoVerse."
Zoe: All right, hi, everybody, this is Zoe. Welcome back to "The VideoVerse" podcast. So for this episode, I'm again be the host, and I have my colleague Alex to be the cohost for this episode. So Alex, you want to say hi? Yeah, right?
Alex: Hello. Excited to be here, excited to chat with Flavio. Zoe, I'll let you keep going.
Zoe: All right, thanks, Alex. Okay, well, he already mentioned that, so for this episode, we're having Flavio from Netflix join us. Really is a great player too. I know you're busy, Flavio, so at the beginning, I want you to introduce yourself to our audience.
Flavio: Absolutely. Thanks for inviting me to participate in "The VideoVerse." It's a pleasure to be here and have a conversation about the world of video streaming.
As you mentioned before you started the recording, there is a lot going on, on the video space, especially live, so super excited to be talking more about it.
Zoe: Yeah, but you accurately got it. This is you're doing some work that I'm supposed to actually process again. Yes, this is a podcast, it's mainly about video streaming, so we like to share a lot of knowledge from our, this time is guests like Flavio, and then to see how the technologies enable video streaming to potentially bring more and more enjoy variable video experience to our end users. So Flavia, I know that you have abundant experience in this, so I'd like you to briefly introduce by yourself, for example- How do you get to this field, and different companies, you have different experience, so just to give us some of your background.
[00:02:35 Flavio Career Background]
Flavio: Absolutely, and yeah, just to introduce me to the audience, my name is Flavio Ribeiro. It's a pretty hard last name for Americans to pronounce it, right? I actually have multiple last names. It's a common thing in Brazil. So for the ones that don't know, I come from a small city in the north east of Brazil, it's called Campina Grande in the State of Paraíba and that's where I started my journey with video.
I was lucky enough that during my undergrad, the government of Brazil was recreating the Brazilian digital television system, and back in the days, the government was highly focused on having the universities participating in that, so that was the first time that I got exposure to how compression works, how bitstreams work, how embedded systems works and all that. So it was super, super helpful for me to have that exposure so early in my career. And then just to fast forward, right when I finished my electrical engineering undergrad, I moved over to Rio. A lot of people are more familiar with Rio de Janeiro in Brazil. I went to work for the largest broadcast conglomerate in Latin America. It's called Globo. Back in the days, live streaming was not really a big thing, but we could already see some signals, that component of video delivery, of content delivery, would impact the destiny of the company.
Zoe: What was the time when you talked?
Flavio: So we're talking around 2011, 2010, 2011, that's when I moved to Rio, and I had the chance to participate on the creation of the first live streaming team, engineering team. I worked for Globo.com. It was internet branch for Globo, so back in the days most of these media conglomerates, they had a internet branch. It was not really part of the like core of the business. It was basically the technology internet branch and that's where I landed. I worked a lot in the backend, as well as in the frontend, still playing with the Adobe Flash technologies- Flash media life encoder and all the services that run with ActionScript, back in the days I was a software engineer.
Right after the workup in 2014, I had the the chance to come to New York City to work for "The New York Times." The company was making a more significant investment in video. The elections were coming up, the Donald Trump, Hillary Clinton battle during that time, so I joined "The New York Times" to help with building a live streaming pipeline. I also had the chance to work on the frontend with player development. It was a pretty great time to work building playback stacks as well. That led me to join, back in the days, CBS Interactive, which was the internet branch for CBS before the merger with Viacom, so before these two companies became what is Paramount today.
Had an amazing time at CBS for about 3 1/2 years where I made a major shift to more of a management, on the management side. We worked in pretty massive projects, including the Super Bowl in 2018 and the Super Bowl in, I think, 2021. And then I was pretty tired of live. I was like, great, my journey in live streaming is over. It was the pandemic. I was like, I just need some time off. And then this opportunity with Netflix came up. It was an opportunity to manage a team that does not relate to live streaming. It was a team called Media Systems, mainly focused in formats and closed captions, fragmented MP4, all types of containers. It felt like an interesting opportunity. I joined Netflix and six months into my new role, Netflix decided to do live, so I came back to live and then that. Yeah, live is kind of coming after me and it's been about two years, 2 1/2 years that we've been in this journey in live streaming with Netflix today.
Zoe: Yeah, basically I was, actually, this is probably many things, so what you just mentioned, but I found out, for example, you're talking about live, right? So here's the streaming you are really involved in live, and then even though, for example, you mentioned that after you joined Netflix, you are not target on that, but then you were pulling, I think you are pulled back to where you were before, and in particular, I think this is very exciting that you mentioned a great series of events about live. So you mentioned that initially in Globo you're doing, at least involving World Cup and that's one of the, if not the biggest, at least one of the biggest worldwide event for World Cup because last year, for example, like a huge amount of audience, they definitely compared that against Olympics.
And then you mentioned that because now we are also at that time about the elections here in the US, and then you mentioned one of the, at least the number one sports in terms of the audience engagement in the Super Bowl and in the football, so all these events that actually involved the live, so based on your experience, how did you like see the evolving of the live streaming offering for all these great events, the technology evolving, the market trends?
[00:08:56 Evolving of the Live Streaming for Great Events]
Flavio: Yeah, those are all very good questions. I was super lucky to have had the chance to work in entertainment projects, in sports projects with CBS, and news projects as well, including the elections with "The New York Times" and CBS News, so it was very interesting to see the differences between the types of content, the challenges that those type of contents bring, not only from the programming standpoint, but also from the technology standpoint and I think as you think through live sports, back in the days in 2014 when I worked on the live streaming for the workup in Brazil, and I don't want to talk about the results and the score of Brazil in that year, but the type of challenges were a lot different than what I see it today.
The scale was a challenge, but at that time the users, the consumers, were not looking at internet streaming as a competitor for over-the-air television, so people would come in to watch anything on their browsers. Mobile phones were still picking up and they all had an expectation that they'd have a subpar experience. They knew that they'd take some rebuffers, that you'd be looking at okay quality, probably not very good for looking at the ball if you're watching tennis, or even soccer, so that was the expectation back in the days about 12 and 14 years ago. Today it's different.
Today, with all the evolution of internet connections and distribution content delivery networks, the expectation is that the sports being delivered over IP will look the same or even better than over-the-air or cable. So discussions around not only the quality, the frame rate, how do we do 60 frames per second, how we do 120 frames per second, how we do HDR, how we do live Dolby Vision, and a topic that is, it's been around for quite some time, but now it's picking up even more is around the latency and the latency for sports specifically.
So there's quite a lot to unpack on this observations, but I think specifically when it comes to sports, you're dealing with an audience that's super passionate about the content they're watching. It's an audience that's willing to pay, it's willing to pay more to watch their clubs, to watch their teams, and they are actually expecting an experience that can fulfill that moment of entertainment for them.
Zoe: I see. You basically mentioned three use cases with segment. I think that's one of the, these just three of the top dominance. You mentioned sports and M&E, as well as news. So particularly I think, 'cause for example, Alex, I know that he has played a lot on soccer, which is I think global-wise is football. And then so you mentioned particularly it seems like used to be the audience didn't expect much for internet stream, right? Though they expect that they should be able to get that from broadcasting like cables and watching that in the big screen on TVs, but now it seems they're not satisfied. You mentioned quite a few, even the numbers, even look for all these kind of things that you just mentioned. The HDRs, the high definition, ultra-high definition with Dolby Visions, and the audience become greedy. There is some technologies that make it happen.
Flavio: Yeah, specifically when you think about, we were talking about 10, 12, 14 years ago, live streaming was basically an add-on on top of what broadcast really did. So you think about the closed captions workflows. The majority of the people on the internet are still relying on 608, 708 for captions delivery that comes from a world or from a time where you didn't really need to deliver five, 10, 12, 20 closed captions language for the same event. The same for ads. How do we integrate your ads workflow?
There's still a lot of lingo and methods and specifications that comes from the legacy broadcast days with SCTE and all that, so the interesting point, like you mentioned, the users are expecting a pretty high experience. They wanna see ultra definition, they wanna see a lot of frames, but at the same time, when you navigate the backend of like how to make this happen over IP, you're gonna see a lot of roadblocks where the technology was not fully redone. It's a bunch of adaptations on how you really deliver a live event at scale.
Even HTTP formats, if you think about HLS and Dash, it can be seen as more of a twist on how to use the internet infrastructure to deliver video at scale by using cached segments. It's almost counterintuitive that you're caching live video, but that's the reality that we had to develop over the last few years. So there's a lot going on. Again, we have to deliver what the consumers are looking for. We have to circumvent those shortcomings on closed captions, on ads, on the IP in general.
And I see some technologies that I consider to be disruptive. I've been following a little bit closer the Media over QUIC what people are trying to do on that front, the evolution of SMPTE 2110 and how to think about the broadcast problem as a more of a software-defined network problem. How do we rely a lot more on the IP technology, but not trying to adjust or remediate the problems from the legacy broadcast, but actually rethinking how to do it all from scratch.
Zoe: So not only just by video signal by itself, right? You mentioned quite a few others like ads insertion that you mentioned, and then you mentioned closed caption, and then also fundamentally underlying the networking protocols that evolving to actually empower the transmissions of the videos. Videos could be huge. Especially you mentioned originally the resolution could be low and then right now huge, and originally there's that limited number of frames per second now we're talking about. You even mentioned 120, that's a lot. We only talk about 60 frame per second for 10 ATP, but then if there is a higher frame rate that you are also associate with the higher resolution and then that's a lot amount of data and that we need the underlying infrastructure to actually support that. And then the internet is happen, and then the video technology is happening, and then you also mention all these others.
Alex: I feel like people also, there's that need for instantaneous like access to everything in general nowadays, but especially with video. Everybody has a very low threshold for buffering or like waiting around in general, so when it comes to an experience like watching your favorite sports team play or like watching your favorite show, you get very sort of angsty is a word I could use for like if it's not, if the playback experience or the live experience isn't up to the standards that we have now today.
Flavio: Yeah, absolutely, and especially now at Netflix, we're just at the beginning of our journey with live streaming. We've been delivering the first events, but we wanna make sure that the technology that we build and the experiences that we deliver to our member base is on pair with what they expect from a service like Netflix. So as a product, as an entertainment offering, we've been building our reputation in quality, in good user experience, good quality of experience, so a lot of the challenges that we've been looking at is essentially how to keep this bar on our live shows so our users are satisfied the same way.
Zoe: Yeah, just like there's so many factors we're talking about, but just now, Alex, you're talking about I think this the latency, and sports latency is definitely important. We don't want, like you mentioned Netflix such has a high bar down there and then if you start to do sports live streaming, the audience want to learn the result instantaneously. But you also mentioned there's a stack technology say even though we are talking about low delay, but there's still some caching that happen. So what is that? Can you just elaborate a little bit more live, how do you resolve the latency, and how that actually relate to the caching technologies underneath?
[00:19:27 How Latency relate to the Caching Technologies]
Flavio: Yeah, that's the $1 million question, or the $1 billion question. I think, as I said before, we're just starting to learn more about the consumer behavior, specifically in the context of Netflix, but looking at what's happening in the industry, there is a lot of different types of content that can be unlocked by reducing latency. We've been seeing a lot of use cases in the wild in different streaming services where betting is becoming a pretty big component of it all, and that actually brings a lot more challenges from the latency standpoint, so I'm still a true believer that depending on the type of content that you're delivering, you might not need that high, sharp, ultra-low latency, but I do believe that there are a lot of other use cases that you might wanna unlock that.
So we've been looking pretty closely on all those evolutions, like I mentioned before, trying to understand if HTTP is the protocol more widely adopted for low latency. We've been following along with the low latency protocols like LL-Dash and LL HLS. There is a lot of other alliances like the HESP Alliance, trying to understand what's evolving in that front, as well as what's happening on the non-TCP world, like I mentioned before. Media over QUIC is still very, very new. You can see that even the most basic specifications are still being developed in the Media over QUIC, but I do think that's a space that we should be looking for.
I think there is a lot of use cases, not only on the delivery side, not only on the last mile side, but also on how can we use QUIC and Media over QUIC to replace some other components of the broadcast chain. So that's how I've been looking at. Some other leaders are also looking at that perspective as well, but as I said before, we're just in exploration phase. It's just the beginning of our journey, so we're still learning quite a lot on that.
Zoe: Right, yeah, I think you mentioned quite a bit like underlying networking protocols. That's a quite a bit of knowledge. But you also mentioned that latency not always means extremely low latency, right? So now, just now you mentioned LL that's low latency, but then there's some times we have to do some compromise between latency and some other factors, especially maybe you mentioned there's got to be smooth transmissions qualities and then there's also, besides the signals, there's other parts.
And so based on your experience regarding the live, what usually if the latency, what's got kind of range and sometimes I think we know that, but could be very low, but that there's some use cases that user will be able to tolerant longer latency, and then with that, there's something that we can leverage that maybe we can add more other content, allowing other tasks to be more relatively easy to implement. For example, news, media, entertainment that may be able to tolerate the long latency.
Flavio: Yeah, and that's a personal opinion. I don't really have much data to back me up, but my understanding from the times where I streamed, I helped with projects on news, talking heads, and even entertainment, if you think about talk shows or talking heads on studios and that kind of stuff, I think there's a lot more tolerance to latency, and I think in those type of projects, you can definitely optimize for other vectors. You kind of glimpsed the trade-offs of latency versus quality versus cost and I think having all those levers well-defined that you can use per title or per show, it's really key.
I think, again, there might be some cases where you wanna push to the edge, you wanna push to the boundaries, you wanna be talking about five seconds or less for glass-to glass latency. You see Amazon Twitch doing quite a lot of that for interactivity. You want streamers to be talking to the audience and that's a use case that latency is a make it or break it for those those use cases, especially around user-generated content. But again, if you go back to an a talking head or a studio show, you might wanna optimize for other characteristics. And I think, like just to kind of tail it all, just to tie it all in this conversation, the biggest challenge with sports is that everyone wants the best quality possible with the lowest latency possible and that's why there's so much conversation about the sports use case because I think it sits in a place that rocks the boat in all these different vectors that we mentioned before.
Zoe: Right, this is actually contradict each other, of course, with the mention that because of the different layers of technology that is evolving, so finally we do observe that the latency will become lower and lower, the quality will become a lot better along the way. And then, but you also mentioned that with the live, there's a courage because there's so many different users, they experience different internet conditions, different devices, and then that's why you actually mentioned that there's a technology behind that, like per show, per title, and then can you also, like we believe that's something to do with all this vary even with one audience, their conditions of the internet connections may vary along the time, and then also there's a huge amount of users I think believe the product and project you got involved, they have a huge courage and with different conditions, right? That has been-
[00:27:35 Different Angles of a Video Delivery]
Flavio: Yeah. And that's something that for specifically for Netflix, Netflix is a company that has been disrupting the industry on the on-demand delivery, on-demand distribution side for quite some time. It's a global company that has to take all the different angles of a video delivery. You mentioned there are some countries that will have a more significant number of challenges from the delivery standpoint. I come from Brazil, so I know how hard it is to deliver high quality content in Brazil. We have another, a few other countries that are pretty strategic for all those different companies and I think that's where some of the challenges are gonna become more difficult to circumvent or to solve.
On the on-demand side, again, we believe that there's more levers from the standpoint that you can spend more time processing the media. You have control over the media. You can do several rounds of inspection on that media before making decisions on how many bits to we spend in that specific scene, how many bits to we spend in the audio side, how to make the right trade-offs, and I think for the on-demand, that's a problem that's way ahead from the research and deployment standpoint so you can see people in Brazil with low-speed connectivity still having a pretty significant, pretty good experience in OTT platforms like Netflix.
And I think for live, again, just to repeat myself, this is just the beginning, at least for us at Netflix, so we're trying to figure out how to leverage and how to build up on all the evolutions we made, how to get all the different concepts that were created or used on the on-demand world and apply that to our live streaming offering. Again, we're just barely scratching the surface, barely starting to understand what's really needed, but we have a lot of expertise, we have a lot of excitement around this journey as well, so I am super, super excited to see how this is gonna evolve not only for Netflix, again, but for the whole industry.
Alex: It sounds like from just like all of your experience in general, there's a real sort of push from all the different industries, news, M&E, sports, to understand the best way at prioritizing those trade-offs. You were talking about obviously for on-demand video, you're able to do that in much slower time, whereas live streaming it's happening on the go, so it sounds like overall there's a push, obviously from Netflix's side, you guys are just getting into live, as you said, you know, six months into when you joined the company they said, "All right, we're gonna do this." So that's still very new, but it sounds like it's also kind of new for a lot of other industries and a lot of other spaces to also look into live streaming a bit more, how to prioritize the different trade-offs to make sure that the experience is best for the end user.
Flavio: That's exactly right. That's what we are seeing. We are trying to focus on the stability of the platform, make sure that we can keep delivering the events that we signed up for, but just trying to come up with ways that we can scale and that we can evolve the platform to eventually be at a place that we wanna be, not only from the latency standpoint, but also from the quality and from the flexibility. So a few things I mentioned before, how can we not only adapt what's in the broadcast, but how can we rethink that? How can we try to approach this problem with a IP-first mindset? How can we make sure that we can leverage the leading-edge technologies to offer the best service possible?
Alex: Yeah, that makes sense. Oh, sorry, Zoe. I wanna just like switch gears completely 'cause you're giving us some phenomenal information, but I just have like a whole question about you grew up in Brazil, obviously, got started in video streaming in Brazil, and then came over to New York, and I mean, so I saw you started at a startup just before you graduated, looked like Ab-a-te, and then when straight at, so you went from like a startup to one of the biggest media and technology companies in the world based in Brazil and then switched it up all over again and went to New York and worked for "The New York Times," so I would just like would love to hear what that transition and experience was like and some of the similarities between media and technology in the US as opposed to Brazil and just yeah, everything sort of that you've experienced throughout that journey.
Flavio: Yeah, that's a very good question. I like to say that I never really had the dream of living my country. I'm a Brazilian by heart. I like samba, I like football, I like my friends, I like the beach. And back in the days when I finished my undergrad, I was focused a lot in Python development and Python was still growing. I think it was Python 2.3 back in the days. And Globo was adopting Python heavily, which really helped me to make this migration over to Rio. Some of the experience with TV and broadcast and Python and some people in the industry that I knew all kind of moving to Globo.
Globo at that time was the place you wanted to be in Brazil, and I think in terms of the differences, when I came to the US, I found a lot of similarities on how the media organizations are defined for Globo. We had the affiliates program, so we had a lot of owned and operated affiliates across the globe. I could see the same at CBS.
But from the technology standpoint, I was surprised to see how evolved Globo was back in the days, especially on experimentation and open source. Open source was a pretty big thing for Globo back in the days. We built the JavaScript player that was used for the World Cup in 2014 in an open source fashion since the first commit, and this player actually picked up and it's still to this day is a pretty well-known player project. It's called Clepper. And here in the US, at least in the companies that I worked for, both "The New York Times" and CBS, they didn't really have a pretty strong culture around open source, so that was something that I tried to influence in some way, tried to bring some of these aspects around open source.
And it's funny that I was just reading an article about how open source is not trending anymore and the industry is looking at open source differently right now and like needing more people to help with open source projects and big open source projects, so it's a world where I think there's some cycles and potentially we're in a cycle where open source is not really in focus right now, but I'm hoping that it will come back even stronger, but yeah, all of that to say it's been pretty wild. It's been a wild journey coming to the US, coming to New York.
I had to learn and expedite my understanding of the language, of the English language. That was a big challenge for me initially and just getting used to the culture. Here in New York, everything is pretty quick. I like to say in New York, like everything is too much when it's like work is too much work, when it's partying, it's too much partying, so it's just, it's a challenge to relate and find yourself and find your place in the city. But after almost 10 years now I can navigate through the city and it's something that's pretty strong in my heart. I don't consider myself a New Yorker, I consider myself a Brazilian, but this is a city that I feel like home as well.
Zoe: I'm happy that Alex changed a little bit, even though it's still very related, so you actually touching open source and then we still feel like, for example, there's a quite open source that also important to us, like a commercial company, commercial setting our encoders and related products that, and then there's FFmpeg, and then also we once talk about Gstreamer and MP4Box. There's a lot of open source tools that very, very relevant and then important to us. Even for the code that we have been doing, there's a lot of open source encoder and decoder in the format of software that always come out, and then indeed it is them who actually pushing forward the new standard to be potentially adopted or paid attention by the community.
Flavio: Yeah, that's a very good point. A lot of the media stacks everywhere, and that's across the globe, a lot of those stacks are highly dependent on open source projects and we've seen some people in the community trying to figure out how to get more people to engage and maintain. You mentioned FFmpeg. I know there's a lot of space for contributions on the FFmpeg side, as well as GStreamer. I've been following more closely the evolution of GStreamer, the adoption of Rust in a lot of GStreamer components and plugins and I'm highly interested at the personal level to what that project can unlock, especially from the live domain, from the live space, like how to leverage a framework that can manage stateful components across different pipelines.
I have a lot of excitement just by seeing the evolution of GStreamer, but to your point, I think we need to have more developers helping with those projects with GPAC as well. It's common knowledge. We at Netflix, we do rely on GPAC for packaging. We do have a lot of interest in helping GPAC as a whole and MP4Box and know the different open source software that comes from that group. But I urge the other companies to be looking at creative ways to contribute and sustain the open source world. I think we depend on that.
It's highly concerning when you see those projects that are massive being maintained by one person or two persons, so just looking from this angle, how can we, as individuals, and how can we as companies help to sustain those open source projects? How can we keep them alive, and how can we make sure that they're evolving and progressing, not only for the industry use cases, not only for the benefit of Netflix or the benefit of other companies, but also the benefit of the community and the benefit of the video and the audio and the multimedia ecosystem as a whole.
Zoe: Yeah, it's all these different aspects or pieces of effort that finally I think are pushing the technology moving forward and then also driving the, I think more and more increment of the video qualities, the user experience on user side. So you actually touch very important aspects of why we come to where we are today. And then there's open source communities, like you mentioned. I think it's worth mention at least, for example, VMAF was the quality metrics that proposed by Netflix in first place, but you're also helping running a open source community to live VMAF at least that actually for the evolving and then that's what we usually recommend to even for our commercial users because they come to, oh some of them even did not hear about VMAF. They say, "Oh, how do we calculate the VMAF?" Then we say, "Okay, there is open source VMAF you can just download." We even come up with some, we are now doing this as down here, but I mainly mention that it's this open source, now only you have good quality VMAF concept down there, but it's leap VMAF actually, just probably get this concept, introduce implementation, then further let people enjoy and experience and play with that and they say, "Oh, okay. This is what VMAF can deliver finally."
[00:42:24 Ways to Evolve VMAF Internally]
Flavio: Yeah, it's a way to move the industry in the right direction, in my opinion. I know there's a lot of other projects built on top of VMAF and looking at different ways to evolve VMAF internally. We have a team that's highly dedicated to like reinventing VMAF and how VMAF can work better for different scenarios. We're talking HDR double vision, we're talking about film grains, like how VMAF continues to deliver value in those workflows on defining quality. Still in the same vein, we've been making a significant investment on the AV1 since the beginning with the AOMedia and how really we can develop and build a codec, compression algorithm and approach to compressing and encoding and the coding video that can really benefit the final users, benefit the content delivery networks and that's all also kind of taking the approach of open sourcing. How can this be built by a community? How can we talk about the different developments? How can we have an implementation reference that you can look at the code base and you can see exactly how that should look like and how that should work?
[00:44:22 Open Source Encoder]
Zoe: Yeah, that's definitely, because you bring out the AV1, there's also some, always I still remember when H.224 AVC was coming out back in 2003, it was the H.224 and everybody would never talk about open source encoder and it was the H.224 open source encoded that really manifest the advantage of AVC H.224. Like I think that standards actually maybe at the beginning pushing the internet video streaming quite a bit and we definitely appreciate unpack one, two before that, but I really, this is unpack four, pack 10, another name for H.224 actually really moving the needle, but then at the beginning people think it's very complicated and hard even though it has a lot of coding efficiency, but it's running really slow, but now access for, with a lot of technologies behind the encoding, it really showed that noise it can really run fast and that now nobody challenge that H.224 is slow. Everybody say, "Okay, look at H.224 how fast it can run." But course, the other factors play that, and again, as you mentioned, this is a community that actually open source played quite a bit of it.
And then I want to actually talking about the codec, I want to also get your opinion, 'cause right now, 'cause we actually as a team know little bit more, not little bit more, we actually spend a little bit more our time on the codec side. There's always a legacy, because thinking about the coverage, and then there's a new technologies, AV1, for example, compared to other technologies, not only on the codec, you also mentioned that on the fundamental like IP looking at QUIC and then Media over QUIC, so what do you think there's some legacy down there because that users are deploying that and now we're talking about new technologies, meaning that there's something that need to change and there's always some inertia down there and how actually you thinking and insights about this and there's a coverage side and then there's a new things that potentially can get more benefit, but then there's also consideration about other things.
[00:47:33 Sustainable Impact by Using New Technologies]
Flavio: Yeah, I think being able to sustain and maintain legacy devices, legacy networks is something that's inherent to what we do. Specifically when it comes to codex and decoders, this has been a problem for the industry for decades now. When you think about the evolution cycle for AV1, we don't talk in months, we don't talk in years, we talk more in like decades how we can really have a sustainable impact by using new technologies and I think that's a pain that we have to deal with, specifically to some deployments including Netflix in this, I think being able to own most of the cycle, if not all the cycles, so all the way from the compression, from the encoding perspective, the packaging, the delivery and playback, that gives some companies, and again, I include Netflix in this, that enables those companies to try out different things in a more flexible way.
One example, AV1 for Netflix is already giving us, paying a lot of dividends for us from the quality standpoint on the on-demand product is something that we have a pretty good penetration already and I think we can benefit from this type of deployments, but when you think about other companies where you rely on third party players and third party content delivery networks and third party software for encoding , that's where I think the complexity is bigger or more convoluted because you have to deal with the legacy in a different way.
So I know it's a long-winded answer to your question. It doesn't really provide a clear opinion of how I think about it. All I know is that it's a pain that's inherent to the work we're doing. We should always be thinking about how long a decision that you make is going to last. So you can see the example of closed captions, the decision that was made in the '50s or in the '60s, is this still the mainstream way of delivering captions? So whatever decision that you're making moving forward, we're now talking AV2, and as we mentioned, it's been a trending topic in this conversation, Media over QUIC, like whatever decisions we're making now, there is a pretty significant chance that that decision will live for decades.
So just keeping that in our minds as we think forward. I know Zoe, offline we've been talking about how to leverage GPU, what are the evolutions on that front? I think that's a lot of promising things happening on that world and whatever decision is being made or whatever evolution is being unlocked by this innovations, we should be certain that they will live for a long time.
Zoe: I was actually getting one point at least from you said, so we have a lot of things, like you mentioned legacies, we need to deal with them, but then basically we bear something visionary in mind and then so it would surprise me you mentioned that not only just month, years, and decades, actually that you mentioned. So as we have something like in the long run and then I think at least we're trending ourself towards that way.
Flavio: I think so and like that's because we depend on the whole cycle of consumer acquisition when they upgrade their TVs, when they upgrade their phones and their devices. And again, we've been talking about AGVC and AV1, but if you look at the numbers, AVCs is still dominant in a lot of deployments so what I'm trying to say is that it's definitely a complex industry that we're not gonna have a way to just magically deprecate everything and create something new, but we should always be pushing the boundaries and trying to figure out how to make the ecosystem better. So GPU, AV1, 2110, QUIC like all those things are intentions in the direction to create an ecosystem that's more flexible, but they will all take some time to mature and to become mainstream, as all the other technologies we've seen in this industry so far.
Zoe: But yeah, as you mentioned, but we are picking up like the innovation mindset and then looking forward to the future. While we're like ramping up, there's still something that we have to mention, which is AI. Still everybody talking about that, especially since later of 2022 that disrupted the GenAI technology came out. So we still want to learn 'cause usually users will ask, "What that will affect the streaming world?" And then what kind of things may be at least step by step, but what is going to be happening in the near future because of the GenAI stuff, and then what happens is going to be in the near future for streaming? We like to hear some about your views.
[00:52:50 Happening because of AI]
Flavio: Yeah, that's a very, very hard question because I have been a little bit in denial mode and not really diving too deep into AI for a lot of other reasons, but I've been following the progress and it's fantastic what is happening in terms of the model evolution. Every time that a new model comes out, new things are unlocked and it's so quickly. We've been talking about changes that take forever to be adopted by the industry, but on the other side, we're talking AI now and I think I've seen AI changing the way people have been developing software, for example, which is a pretty like transactional upgrade or progress. If you ask me, developers that never worked in a specific programming language, they're now able to build software in this language and I think those are minor non-disruptive benefits that AI is providing to us.
But I'm kind of trying to look at like other ways that AI can impact our daily routine, not only on the media and entertainment or on the streaming side, but also how we go about life, looking at how the government makes decisions about, you know, their countries or the cities or the states and how AI can bring a perspective of what worked best in the past, but specifically to the video streaming, the video industry, I think the altering of content might change. It's not something that I've been heavily involved in the creation of the content itself, but I can already see a lot of glimpses where AI is transforming how people expedite storyboards, how people expedite the different types of content being created.
But yeah, it's a very, very hard topic. You kind of have to be closer to have a stronger opinion about what really this technology will unlock. You know, there's an analogy. When you're a fish and you're in the water, it's hard for you to learn the concept of being wet. That's kind of how I feel it, you know, I mean, I'm embedded in technology so it's hard to understand like how to zoom out and understand how AI will change my own world and my own day-to-day job, so I will leave it there. I could talk over and over about the things that I've been thinking about on the AI side, but I might just not be realistic, I think, about some of the things that are going through my head right now.
Zoe: Well, thank you for actually the analogy you just mentioned. I think we're all sort of some kind of fish in some kind of water at this moment and I even though we all want to actually jump outside to actually looking into from a bigger picture and that's always the case, but as long as we are aware of this, we are aware of the limitation down there, there's always we can make better, make further progress.
We have been almost an hour and then so, but actually I, not only at is from my side, not only learned about your sharing, I think what I learned is even though, especially from the last thing that we just mentioned, we're aware of our constraint and limitations, but I heard someone always mention that it is the actions that actually produce information, meaning that we don't know something, but as we move ourselves forward, learn new things, act new thing, contribute new things, doing new things, we start to learn more and knowing more along the way, that's what we learned based on all your past, based on what you are sharing.
Flavio: Yeah, I think the quicker you learn how to learn, the better, especially in this industry. If you learn what makes you a quicker learner, if that's reading, if that's watching, if that's producing, trying out, I think once you find your best way of learning things, that's how you're gonna be unlocking most of your talent. You're gonna be successful not only in your career, but also in your personal life, being surrounded by people that are also learners. I think that's one thing I've seen here in New York specifically, just being surrounded by people that are always hungry for learning more. So if you can find your way of learning, the quicker, the better, I think you will navigate your journey of life.
Zoe: Well, really thank you. And I still want to have one comment because you always mentioned you treat yourself as a Brazil person even though you are in the big city of New York, but you already take New York as at least another home of your life, even though it's too much, but that gave you the opportunity to expose yourself to a lot of new things and information, different categories, that's actually really have yourself, have the opportunity to learn new things, become even a quick learner in that environment.
Flavio: Yeah, New York transforms yourself in so many ways and I'm true to that. I did, you know, I did transform myself multiple times. I did become a different person, but I'm a Brazilian by heart. I try to mix and match the things that I like the most in Brazil with the things that I liked the most in New York and I try to go about my life one step at a time, one day at a time, and just trying to enjoy the best moments of both.
Zoe: Well, thank you. And then I'm going to actually close this episode and then I hope everybody, really at least myself, actually, take something from here and then we enjoy your journey and then we're looking forward to learn more from your future. And then I will definitely pay attention to your own pursuits as well, so what's going on in the whole world in the streaming, as well as in the latter world regarding for potentially on AI and all the other things because the world is changing fast, but then we all just have the belief it will only change better along the way. Thank you.
Flavio: Yeah, I appreciate it, thank you. It's been amazing to have this conversation with you both. It's a pleasure to participate on "The VideoVerse" and yeah, let's keep in touch. I promise you when I learn more about what AI will unlock, I will come back here so we can have a special episode just talking about AI and video streaming and the media and entertainment world. But I had a great time. I appreciated the time and I hope it's been helpful and fun for the listeners.
Zoe: Well, thank you and also thanks Alex to be my host, co-host. We also thanks everyone to paying attention to this episode and Flavio, as you just mentioned, we're looking forward to have you back down the road.
Flavio: Absolutely, you can call to me.
Zoe: Thanks, everyone. Thank you, and then I'll see everybody the next time.
