NETINT Technologies about a Video Platform used by 97% of US Football Teams

NETINT Technologies about a Video Platform used by 97% of US Football Teams

Show Video

Well, it's 8:00 AM, on Pacific Time anyway, so  on the West Coast, I guess. Hey, welcome to this   edition, another very, very exciting edition  of Voices of Video. Casey, are you ready to   have an awesome conversation? Absolutely. Let's do it.  All right. Well, good. Well, let me  introduce you. So as everyone knows today,   I'm going to be talking with Casey Bateman, who  is with Hudl, and you're going to hear all about   Hudl. But we were just chatting, Casey and  I, about how the company was founded. I'll  

let him tell the full story, but they're  actually one of the OGs in streaming video,   and I have been fortunate enough to have been a  part of, not Hudl, but some other companies in   the same vintage. It's amazing how on one hand, 15  years ago is 15 years ago, and yet it seems like   yesterday how far we've come in streaming video. Casey, welcome to Voices of Video. Thank you for   joining me. This is going to be great. Yeah, thanks for having   me. I'm excited to be here. Yeah. Cool, cool. Well,  

let's start, obviously give a quick intro to  yourself and tell the viewers what you do,   and then talk to us about Hudl, what  Hudl's building, and let's get into this.  Yeah, absolutely. Yes, I'm Casey Bateman. I'm a  principal engineer in the office of the CTO at   Hudl. Specifically what I do is I look into our  video technology stack. I kind of own that area,   technically speaking. My job is to look at new  technologies, assess the landscape, and to try   and plot out a strategy for the company around  our video vision and how we're to execute on that.  It's a big job plotting the landscape and the  vision of the future. But tell us about Hudl,  

because you are a remarkable company,  especially... Well, you're, I'm sure,   a remarkable company, but in the area of  streaming. I have to believe that probably   not all of our viewers even know who Hudl is, and  yet, I think everyone's going to be shocked when   they hear the scale of operations that you  run and what you do. So what does Hudl do?  Hudl is... We are a technology company that  focuses on empowering coaches, athletes,   and fans' sports experience. We build a coaching  tool platform that really helps coaches and  

athletes work their game and get better at it.  We have a platform then also for our fans who   can't necessarily be at games, or are at games and  want to enrich their experience at those games,   or even just follow their favorite athletes and  teams, just watch that content. We're like a full   service sports video suite of products. Yeah.  Yeah. You were saying OG Company. Yeah. Company  started back in 2006. Got our first customer,  

was actually in 2007- Incredible.  ... and stuff, with the University of Nebraska. Wow. Wow. What's the origin story? Are the   founders athletes that were trying to solve  a problem, or technologists that were just   looking for an application? Both? Neither? Yeah. Well, actually, yeah, our three founders   is Dave Graff, John Wirtz, and Brian Kaiser. They  were attending the Raikes School at the University   of Nebraska, which is an entrepreneur computer  science program, the stuff that the University   of Nebraska has. Dave Graff, who is our CEO,  was working with the football team at that time.  Just to paint a picture of how it was back then  for you, these teams, when they'd exchange video,   they were doing VCRs. A lot of time  they would have their video recordings,  

they would cut everything and record  to a VCR, and then meet other coaches,   drive and meet them halfway. It was- And exchange tapes, right?  Exactly. Literally.  Exactly. This is what it was like, and it  is hard to think, based on where we are now,   there was a past in which that was the thing to  do. Dave Graff had the idea at that point in time,   and John Wirtz and Brian Kaiser joined in, and  they developed a product they called Huddle. Now,   it was spelled a little differently back  then, like the actual, what you'd think of   when you're talking a huddle, was the spelling.  It was a smoke and mirrors product, that's what  

we call it back then, that they took to Coach  Bill Callahan at that time, for the University   of Nebraska, and really sold them on it. So for that first year, they were using   this new product, or Huddle at that point in time,  to test it out with the school. And then in 2007,   in June 2007, we officially signed the Huskers  as our first client. It took about another year,  

until about March of 2008, until we finally got  our second client. If you know the football story   behind Coach Bill Callahan, he wasn't with the  Huskers very long, but then immediately after   went to the Jets, and he brought Huddle with him.  So our second ever client were the New York Jets.  Professional team. Incredible. Right.  Incredible. Right. So that was the idea, originally, is   Huddle was going to be made to be targeting these  college and pro teams and stuff. So for that first   two years, we had a lot of interest from high  school, small college teams, wanting to use the   product. It was one of those things that, up until  that point, we just turned them down. But in 2008,  

decided we needed to shift, split the product  into two sets, the whole Pro Suite, and then Hudl,   and start tackling that high school market.  And from there, Hudl just started exploding.  Incredible. Yeah. By the end of 2008,   we got an additional 12 high school clients.  In 2009, we got up to 350 in total clients. 

Wow. Then by 2010, we were up to 1300 high school.  Wow. Wow. That's hockey stick adoption.  Was it just purely viral? I guess-  Yes. ... it almost  

makes sense. You talk about network effects.  If two coaches were driving to meet, all it   takes is one of those coaches to adopt it and you  automatically get the other one, right? Because-  Exactly. ... they'd be like, "Wait a second,   you don't have to go drive..." What a pain. That  could burn a Saturday afternoon, or whatever. 

Absolutely. It burns your time. A Monday night, or whatever. Yeah, wow.  Especially these high school coaches. They want  as much time as they can doing what they love,   and driving back and forth is  probably not what they love.  Yeah, incredible. One of the statistics that I  think we put in our poster art for this particular   interview, was that you have 97% of US, and I  guess we should say American football, right?  Right. We are going to get in to talk   about the technology. So anybody who's listening  and saying, "Wait a second, are you going to talk  

about what they built?" Yes, we are. But I think  sometimes, as technologists and engineers, just my   observation is, we're sometimes almost too quick  to dive in to talk about the nuts and the bolts   and the protocols and the standards, and without  the context of the application, how a product's   being used, why something is built a certain way,  sometimes it can either lead you to not realize   really the full breadth of the significance  of why something was built the way it is,   or you may be asking, "Well, why are they using  that standard, that technology? That doesn't make   sense." Well, when you understand the devices, for  example, then you're like, "Oh, okay, I get it."  So that's just my caveat to say, bear with us  here if you're looking to get quickly into the   nuts and the bolts, we're going to do that.  But you have 97% of American football teams,   so that is across high school and colleges, the  high school, college pro? What exactly is your   addressable market there that you look at? For the American football market, yeah,   it is 97% of the high school football teams  in the US are all using the Hudl platforms.  Amazing. Really, the ones that aren't, are generally when   you start getting into the six man football teams- I see. 

... like smaller schools. Small schools, yeah. Yeah. Wow,   that's incredible. And then, you have schools  and customers outside the US? Or are you-  Yes. Regionally, where are you?  We are actually global. We have a big,  big presence in the UK area. We have   a big presence over in Sydney, Australia. I hear they also have football over there,   but the ball looks a little different. They do. They do. It's a little different,  

and they're not carrying it as much. That's right, that's right. Yeah.   Hands not so much. Hands off, right?  Yeah, that's right. That's right.  I guess they also use their heads,  

although I think I'd rather use my head  to hit their ball than American football.  Oh, yeah. No kidding. Using my head to hit another large person.  Yeah. Yeah. Okay.  We've definitely expanded. We service over 40  different sports right now when we make up Hudl,  

and there's over 230,000 different organizations- Incredible.  ... and teams. Yeah, we are global- Incredible.  ... for sure. Yeah. Yeah. Absolutely   amazing. Well, we have talked through the  years, and so I've known about Hudl and worked   with Hudl for a little while, but until that  point, I too was just absolutely shocked at,   how is it possible? Being around the industry  a little while you can sometimes feel like,   oh, I know everybody who's built the significant  platforms, and then Hudl pops up and you go, "Wait   a second, how have you guys been hiding out?" Yeah, congratulations. Now, you were not a   part of the original founding  team, right? You joined after-  Right. Yep. ... the first platform was built? Okay. 

Yep. I joined in 2013, right at  the beginning of the year there.  Okay. All right. But you're telling me that  you were scaling. 2010, you had, I think,   1300 schools. So clearly, 2013, well, let's  face that's 10 years ago now. So you've been   there through some growth. Why don't you tell  us what the platform looked like then? Just  

architecturally, how did you build it? Were  you building on the cloud? Did you own your   data centers? Talk to us architecturally, and  then maybe walk us through what it looks like   today from an architectural perspective. Yeah, absolutely. Thinking back to that   2010 time range, Hudl, we mainly targeted  stop-and-go sports, or clip based sports is   what we've referred to them as, or that's kind of  like the American football market. That was our   bread and butter. Still is our bread and butter,  even though we've got so many other products.  Back in 2010 through 2012, Hudl was clip based,  and I say clip based because most of those sports,   it was, like I said, stop-and-go. When you're  looking at a football game, it's anywhere from   a five-second play to maybe a 40-second play  for one of those really long time periods.  That's a really good point. I remember you made  that comment to me at one time when we were  

talking, and I went, "Yeah, that's right."  American football, in particular, I guess   you care if the running back is repositioning  themselves and whatever, but let's face it, what   I really care is once the play starts. There's a  significant amount of time where the play hasn't   started, players on the field and whatever. So  that's what you mean, just to be clear, about-  Correct. ... the clip based? Yeah. 

Yeah. There's a lot of dead  time that coaches really don't   want to see. They want to see the action. Yeah, that's right. Yeah. Again, they're busy,   so they don't need to sit through all of that. Right. So at that point in time, our video came   in individual clips. We used WMVs at that  point in time. We used WMV because it played  

really nice with Microsoft Silverlight, if  you remember the Silverlight Technologies?  Yeah. Yeah. We used Silverlight back then   because it allowed them to build a pretty rich  experience where we could actually draw directly   on the player canvas. We could do more intelligent  clipping and stitching of video files together,   which is an integral part to the coaching  workflow. They need, at certain point in   times, much like if you're watching an ESPN  broadcast and they're breaking down the film,   they need to telestrate, essentially, on the video  play, and draw out where motion should be, where   routes need to be. We used Microsoft Silverlight  at that point in time with WMV files. We started  

shifting into MP4s quickly after that, as MP4  became the streaming standard, and was definitely   the streaming standard at that point in time.  Yeah, so most of our platform at that time was   all online. Actually in 2009, we signed with  AWS. We were very early adopters of Amazon Web-  2009. Wow. Wow. You're an OG on AWS as well. That's right. That's right. 

Yeah. All of our video files were then   hosted. We host our video files on AWS and stream  them down through the Hudl.com website. Yeah,   at that point in time, Hudl definitely looked  very different. It was that bright orange color   that you would see almost all over the place  and was really designed to be optimized for   a football coach's experience. Spreadsheets  are almost like a spreadsheet or table looking  

area where they could enter in play-by-play  data, as what happened, what were results,   and then you'd have your video player Chrome. The original experience was all in the browser,   and I also was a part of building some demos  with Silverlight back in the day and showing   some very cool interactive features. It's  interesting you bring that up. I can remember,   I was a part of a consortium called DECE. We were  doing the digital locker in the original, original   days of transactional VOD. If you purchased a file  on one particular platform, or if you rented it,   not only was it only viewable on that platform,  so like Voodoo, you couldn't watch it on, well,   back then, let's say Apple, but you couldn't  watch it on some other competing platform. But  

you couldn't even watch it on a different device  on Voodoo. So if you originally transacted it,   purchased or rented, whatever that transaction  was, on your connected television, which way   back in the day there were like three, but  you couldn't then watch it on your laptop.  I can remember some very, very cool demos that  Microsoft did with various partners showing   just some really cool... basically how you could  take the interactive DVD or Blu-ray experience   and translate that to streaming, which again,  I think a lot of us forget. Even though now,   it's a little bit archaic and resolutions are  very, very low, but there were some pretty cool   Blu-ray and DVD experiences, interactive  menus and all kinds of really cool stuff.  That's an aside though. You were  streaming into browsers. Now, today,  

what is the dominant playback mechanism?  You must surely support and have very strong   support for mobile devices, I'm guessing? Actually, playback is still split between   the different devices almost perfectly in  half. I just looked up the numbers here-  Oh, interesting. ... to get some updates.   I would say half is still web traffic, and we  have various protocols we stream over, MP4,   HLS, for varying platforms. And then, yeah, we do  also have our Android and iOS applications where a  

lot of our athletes utilize those to watch their  video and do some of their review. We're getting   into the fan streaming space more and more, so  we have a whole host of smart TV apps as well.  If someone uses one of your apps, do they  get some additional functionality? Or is it   just convenience, not needing to use a... and  of course some devices, maybe a browser's not   the best playback mechanism. I'm thinking  connected television, as an example, but- 

Yeah, using our apps, and it depends  on which app they would use, but yeah,   they would get some additional features  and availability. For example, our iOS app,   we had the idea that anybody has this  phone with them at any given time can   capture different moments in sports. That's  a great capture mechanism for getting video   into Hudl. So we do have capture in our  iOS and Android applications as well.  The only sports that I played were... I was  in track. I was pretty competitive in high   school and ran a little bit in college, so I  can't exactly relate. I never played football,   never played basketball, but I can imagine a coach  standing on the sidelines now, everybody has an   iPhone with them or an Android device. You could  whip that out in theory, I'm just assuming this  

is the function, and capture 5, 10, 15 seconds  of something, and then you upload it, right?  Yeah. And then it's   available. Yeah. That's interesting. Wow. Okay. Well, a lot of streaming conversations   are in the distribution context, and that is  the one to many, or the one to a very large   number of manys, meaning, certainly for live  events, like a Super Bowl, if we're going to   stay within the sports conversation. Every year  it's like, "Well, how much larger was this year's   audience than last year's for streaming?" But  it seems to me, based on what you're describing,   your use case for the platform, from a platform  perspective, is different. Do you run events  

where you might be streaming to millions of  people a single event? I'm assuming you don't.  Yeah, not really. We don't really have  those big events. Most of the time we're   streaming just directly to teams' feeds. To  their fans, to their athletes and coaches.  But on a Friday night, for example, especially if  we're talking high school sports, there must be a   peak time where all of a sudden your platform  is pretty crowded, right? Because that's when-  Absolutely. ... a lot of schools   are playing. There's even a time of the day that  you can look at from East Coast to West Coast,  

if we're just going to talk about the US. That's  a different type of challenge. And believe me, I'm   not suggesting one is easier or necessarily harder  than the other, because it is a challenge when you   have a stream and it's a single stream or a small  number, but it's going out to a million people,   okay? There's scale and there's challenge in that.  But it's different when you might have 10,000   simultaneous streams that individually might just  be going to a very small number, but yet your   platform perspective, you have to power them all  equally, and you have to provide a great service.  Can you give us some insights into how  your platform is built, and maybe even   some surprises that someone might say, "Well, why  did you make that choice?" And then you're like,   "Oh, well, because I don't have to deliver to  a million people, but I do have to be able to   run on a Friday night," whatever, "X number of  hundreds or thousands of streams reliably without   question." So yeah, give us some insights of... Yeah. Absolutely. Yeah. You hit the nail on the  

head there earlier when you were talking about  we are built definitely different. Most of the   time our content is coming from teams and coaches  and is targeting their athletes, their fans. So we   will have a recording, these two hour long events,  they may have a couple hundred views on them. So   we're not talking big numbers, which definitely,  when you're talking about the distribution side,   the number one thing you might immediately think  is, yeah, our caching strategy is very different.   Where a lot of these sites, they may hit for  really high cache rates, I'm very happy with   a 40 to 50% cache hit rate. That's actually  fantastic performance, in that situation. 

But Disney streaming would be like, "Oh  no, we have a serious problem here."  Exactly. Exactly. So the things that we have  to optimize for in that situation are going   to be regional optimizations on trying to pull  content down to the users. We look at things like   pre-caching content, bringing content, either when  we're talking our HLS streams, bringing fragment   by fragment content down to the users, just as  they're watching the stream, we know what the next   chunk of video they're likely to watch is going to  be. We try and get that pulled into these PoPs in  

the regions where they are watching these videos.  So pre-filling of the cache is a big thing for us.  Also, of a diversity. I know this is  important across multiple organizations,   whether or not they're getting millions of  views on a single stream or not, but also   the diversity of PoPs available in diversity  of CDNs. We run a multi CDN environment here,   and that's because we don't usually have isolated  chunks watching the same stream across multiple   areas. You might have isolated chunks of users  watching our content, but then across the state,   someone's not watching that content. We need to- Yeah, understand. 

Yeah. We have to target, essentially, our users  for what's going to be the best performing CDN   potentially for their region, for their  exact ISP and where they're accessing   from. So how we target our users, what CDN  we provide to them, is slightly different   than what I would say you might see out of a  YouTube or a Disney, or something like that.  So what I hear you saying, and I don't want to  put words in your mouth, so correct me, but almost   always when you're talking about these really  large scale events, where it's streaming a single   stream to millions, it's about cost reduction. So  very often the CDN decisioning gets down to... and   it is splitting fractions of pennies, because it's  meaningful when you're delivering to millions.  

But I didn't hear you mention... obviously you  care about cost. I'm not suggesting you don't,   but is it true that your decision is not strictly  necessarily based on cost, and it's performance,   as you stated? Or it's even just purely about  coverage? Like if one CDN doesn't have coverage in   a market, well, you need someone who does, though? Absolutely. Yeah. I would say, yeah, cost is   always going to be important. You're right. But- Of course, yeah. I'm not   implying that it's not, but- No. Absolutely. But yeah, mostly it's just the   availability of the network and the performance. Yeah. Interesting.  It is the more important thing. If a coach  can't access their content because of a  

couple regional PoPs, or one CDN are out,  then you can be assured that our support   team is going to have their phones ringing. Going to get a trouble ticket filed. What is   the role of quality with this content?  I'm asking the question because really,   I don't know. Is this one of those shocking things  where coaches care about quality even more than...   or is there a different bar of acceptable quality,  therefore you can encode at maybe lower bit rates?   Are there some sports that are more perceptive  to quality than others? How do you approach that?  Quality is really important in the sense that  there are a few things that we have to get   absolutely right in encoding our videos. That's  going to be the field details and jersey details.   One of the things that you don't really- Oh, good point.  ... think when a coach is breaking down content,  you'd like to think that... A coach knows the   names of their players. Obviously they know the  names of their players, but when you're watching   a video and you may be getting a video shot from a  press box on a football field, everyone's wearing   a helmet, all you're seeing is jersey numbers.  You need to be able to quickly identify who did  

what and where. Clarity on numbers, and one  of the things we see is with a lot of those   lower bit rates with some encoders out there  on the market, some of those finer details,   the jersey lines and stuff like that. Also  with hair, you get people with longer hair,   it blocks jersey numbers, and the encoder creates  these artifacts or unnecessarily blends together   jersey numbers, and features just together, and  it makes it more difficult for coaches to see.  So, yeah, we do generally have to provide a  slightly higher bit rate or use an encoder that   is going to be optimized for providing those clear  lines, and also bolder colors. One of the things  

that's really easy when you're talking about an  outdoor game, is getting that color blending.   Some encoders work really well at keeping your  colors bold, not desaturating content too much.   Other encoders may not do as great a job. To take this back, quality plays just a   really big part of the sports video experience,  because when we're watching content, we use the   identifying characteristics, like jersey number,  field lines, to help tell the story of what's   happening on the field. Interesting.  Right. So those- Yeah. Makes sense. 

... things have to be clear and easy to read. Yeah. Yeah. Yeah, that's interesting. That   completely makes sense. So what is the  architecture? What I hear you saying is, is   that a large percentage, if not nearly all of the  source material, you're actually not generating.   Are you putting cameras out in the schools, or is  it on the school and the coach and the players to-  We started off, it was on the schools and the  players. We have, over the past several years,   been developing an in-house  camera. I've actually got- 

Makes sense. ... my development   rig of it right here. It's this- Hey, cool. Cool. Let's see that-  You can't really see... There you go. There we go. There we go. Okay.  Yeah, there we go. Yeah.  That's our- Cool. Cool.  ... Hudl Focus camera. Basically, it's a  stationary camera that gets set up in gyms,   gets set up on football fields in different areas  that will record. Uses a collection of sensors,  

maps them all together in a 3D environment,  does player tracking and other AI to help track   the ball. And we create smart tactical feeds of  video that is uploaded directly to Hudl that way.  Interesting. Okay. But overall, video can come   from a coach from their iPhone, from an expensive  camera. A lot of these Texas schools, they have   nice expensive rigs and stuff they love using- They love their high school football in Texas,   don't they? Yes, they do.  By the way, I grew up in Colorado, and I grew up  when Nebraska and CU were the absolute powerhouse   teams, so yeah. Yeah. So I know a little  bit, even though I didn't play the sport,   I know the culture. I can relate to it. Yeah, it  actually is really amazing. If someone's not been  

exposed to it, just go to a high school football  game in Texas, or back in the day in Colorado on a   Friday night, and it's quite an experience. It is.  They take it serious. It's a way of life for them.  Yeah. Yeah, totally. Okay. So you have content  coming in from a lot of different sources,   many of which you don't control, meaning that  it could be very high quality, very well shot,   great lenses, great cameras, all that stuff, and  others, it's a phone and a shaky one at best. 

Yeah. So then, are you doing some processing   in the cloud? Do you transcode that? What do you  do with that video once it's on the platform,   and do you have a decision tree? Are you using  some sort of, even, I've got to get it in there,   AI, to maybe do some sharpening or some  processing or something, to improve quality?  For most video, the way that we've operated  for so long is we've taken the single encoding   ladder approach. We didn't have a smarter way  to provide more bits to a certain stream. It's   just historically how we've done it. And some  of that is around cost management. When you've   got all these streams, we've got 100 petabytes of  video out on S3 in our video library right now,   it's a lot of video. It's a lot.  And in order to keep our costs down, but  also provide the best possible experience,   we generally just test to find what is going to  be the optimal bit rate in which 95% of our users   have top quality for the highest renditions that  we have. And that's historically how we have done  

things. Now, we are looking at exploring a little  deeper, maybe providing different experiences and   stuff. We're exploring, obviously, some  of these new codecs that are out here,   like HEVC, AV1, do a great job at providing  that better quality video off the bat at   those lower bit rates that can- Yeah, lower bit rates. Sure. 

Moving forward, I think we're looking to  optimize, possibly on a per sport level, which-  Interesting. ... we talked a little bit earlier. Yeah,   there are some different requirements for  different sports that might just have a worse   experience. And thinking, water polo is a big  one. Like most people in the video world, you've   got the moving water. There's so much motion. Yeah, water is super challenging to encode.   All those pixels are in motion, continuously.  There is no step, by very definition, with water. 

Right. I know one of the things that we've  had to do in special situations like that,   is we have to slow down the encoder significantly.  In those situations, you have to pay for a little   additional compute time on that, and then,  yeah, you have to up the bit rate on those   streams to provide a much better experience. Are you standardizing on 1080p for what the   coaches and the players are looking  at? Is there a need to deliver 4K?  Yep. It depends on the sport and the level  there. The NFL, they still run on 720p.  Isn't that amazing? It is amazing. However,  

our coaches, most of them, they want 1080p.  Our football, we are delivering 1080p.  Yeah. I have to ask a question here, and you can  answer it as sufficiently general as possible to   not give away any secrets or inside... but  it still is shocking to me how broadcasts,   and you mentioned the NFL, and not  singling out the NFL because it's true,   major leagues go around that are running  on... sometimes there's even MPEG-2 happening.  Now, I'm talking more in the broadcast  distribution chain, but my point is,   is where there's literally two, three, sometimes  four generation old technology that they're still   using today. 720p or interlaced. It is shocking.  People just don't know how much interlaced is   still happening even in live TV production.  Now, interlaced, that's largely, I think, gone,  

but three, four, even five years ago, there was  still a tremendous amount of interlaced content   that had to be then deinterlaced. Meanwhile,  everything's being streamed at 720p, or 1080p,   or 4K, all progressive, and yet the production  chains are so old. Why do you think that is?  Well, yeah, it's funny you mention that. I'm asking just a general question. Why   is that? And 720p at the NFL. Why is it  that a high school team in Texas can have   better quality than the professional league? Yeah. Real quick. It's funny you do mention  

that. Over 50% of our video we get in, we  do have to deinterlace as well. So yeah,   it's still out there for sure. Wow, amazing.  I can tell you that generally, when you're  talking about these bigger programs,   I'll speak specifically to the NFL, there's  just the concern around security. Security,   it's [inaudible 00:38:37]- Oh, interesting.  ... just more worried about- So their view is the lower the resolution,   the less useful the video is? Is that- Well, not necessarily. See a lot of them,   they actually still record to SD cards and they  have runners that will run video content from   the sidelines up to the press box and stuff. Some  of them are under the belief that going to 1080p,  

they would fill up those significantly. A lot  of these hardware devices that they use to   record to SD cards only support 720 feeds. Crazy.  Yeah. They're- So it's like literal   hardware limitation that could be as simple as,  I have a file here, everything will play it back   except for this one device, and therefore I have  to reduce my experience because of this one single   device in this somewhat complicated distribution  chain? I have to say I'm not surprised by this,   because again, getting back to the interlacing  issue and other... in my past dealings with   broadcast, whenever I've asked... like around  4K, that's where I really became exposed. "Why  

is it that 4K is so hard to adopt?" And  everybody says, "Well, it's the cost."  It actually isn't the distrib... Yes, there's a  distribution cost associated, but it's the cost   of supporting throughout the entire pipeline. Correct.  And it can be as simple as my graphics... box.  I'm blanking out. They have a name for it. But   that it doesn't support 4K, and there we have  to go through all kinds of gyration. We have to   basically convert it from 4K to 1080p, and then  we upscale and it's like, "Well, now we lost it,   so we just run 1080p all the way through."  And you're like, "You mean just because of  

the graphics overlay device, that hardware,  the consumer's missing out on this beautiful   experience?" Like, "Yeah, that's it." Wow. Yeah. And that was the other reason, is cost.   Cost is the big thing there, especially  jumping to 4K. When you're talking 4K,   a lot of these boxes that these coaches have  in their office that they're watching on,   they're not going to support the latest HEVC.  They're going to be 1080p televisions and- 

Yeah, it's H.264, and yeah, it's an old TV.  Yeah, interesting. So you mentioned AV1 and HEVC,   and it's certainly of interest, and we've had  conversations about those codecs. Platforms have   different strategies for adopting, shall we say,  next-gen codecs. One strategy can be, "We're just   always going to use the latest and greatest." And  then there are the more conservative platforms,   are like, "Yeah, we're not necessarily going to  be the first to adopt." Where are you in this  

continuum? And then tell us why. Because, again, I  think the why is important. It's one thing to say,   "Well, we're pretty conservative,"  but then, is that because people are   on devices that just simply don't support AV1,  as an example? Yeah, but tell us where you are.  Yeah, absolutely. Yep, you're right. The simple  answer is yes, we're going to be a little   more on the conservative side of this. Why that  generally is, is when you look at our user base,   our power users are going to be coaches, and you  have a big mixture of the types of coaches and   the technologies that they're comfortable with. When you look at our user base, where they're  

running on, whether or not it's personal devices  or even school devices, keep in mind we're looking   at high schools and stuff, and a lot of public  school systems aren't going to have access to the   latest and greatest technologies out there. Yeah. Yeah. Yeah. Yeah, that's true.  Generally we got to look at what devices are our  users accessing our content, for what platform   they're accessing as well. I think when you start  talking more of our fan content, we get to branch   out beyond the high school environment and start  looking at just device share around the world.  Yeah. When you start, once again, getting   into our core platform, yeah, we're dealing  mostly with school computers and stuff like that.  Yeah. Yeah. Interesting. Yeah. Generally we are, like I said,  

more conservative on that. We are getting into  the HEVC world, given the Chrome announcement here   last year about supporting HEVC playback- That was good for you, I'm sure.  That's really big for us. Since the bulk of  our users, about 98% of our web traffic is   going to be either on Chrome or on Safari,  both of which will support HEVC if the   hardware capabilities are there. That's right. That's right.  Currently, we're sitting at about...  I double checked it again last week,  

we're just about 71 to 72%, right in there,  of our users have that hardware capability-  Amazing. ... to decode HEVC.  Amazing. Wow. Wow. How do you know that? Yeah. What we are doing, we ran a couple   tests. So really fun thing you can do for anybody  wanting to [inaudible 00:44:39]. There are a   couple HTML5 video APIs you can say for, can play  video codecs. I have to look exactly what it is,  

but essentially, by specifying your codec tag,  you can determine if the system can properly   build the decoding chain, connect the pins, can  decode it and stuff. So that is based on whether   or not Chrome or Safari tells us that they  absolutely do have support, or is most likely   supporting this stuff. So definitely support- Got it. Got it.  ... gives us approximately about 65%, and then the  additional, there's about 10 to 15% there on top   of that, that are in the some support. We assume  that about half of those actually can support. 

Actually can, yeah. Yeah. You mentioned that you  have a custom app across certain platforms. Do you   put a software decoder in there, or would you  contemplate that? Or are you 100% dependent on   just the hardware support on the device? Right now, it's going to be hardware,   entirely dependent on that. And some  of these newer codecs, we've looked at   potentially even LCEVC- Like AV1, for example.  ... or AV1. Yeah. Oh, LCEVC. Yeah.  Right. And putting those in there. And just for  our users and the time they're spending on our   app, we don't see as good of a battery life.  That's going to be your big thing that you   see when you're putting in a software decoder on  these newer products, is battery life really takes   a hit and stuff. It hasn't been worth it to us,  especially when you're weighing at the development  

cost to put something like that in, with the, all  right, if the majority of our users actually need   this for playback, does it even make sense for us  to encode a quality of that to them? So generally,   when we are looking at new qualities, new codecs  to use, it's dependence on hardware support,   or a codec that is not going to sacrifice  battery life significantly in mobile devices,   are what we're going to have to look at. I know some of the... like dav1d for AV1,   is really, really power performant, which is to  say that it actually is amazing what it can do   and how little energy consumption it needs. At the  same time, there's always that trade-off, right?   You still have to push the user even to upgrade  your apps so you can build all this, and then if   an insufficient number of users actually upgrade  their app on the device, then they don't even have   access. It's a lot to push upgrades even. Yeah. I would say even then though, we do have really  

good adoption rate of our new app versions when  they come out. We're generally looking at about-  So, people are... Yeah. 15 day to a month to the   majority of our users getting updated apps. And the majority is what? 90% of the base?  Correct. Yeah. Yeah. Yeah.  So, [inaudible 00:48:12]- Which is good. That's really good.  AV1 has been something we've actively  been looking at since Apple announced   that there was AV1 support with the HLS spec. Yeah. We didn't talk about DRM. Is DRM required  

for any of your content or on the platform? Not currently, no. It's something that we   haven't heard. We're waiting to hear back  from our coaches on whether or not that is   actually required. I would say it's not likely  going to be something that we will ever need.  Yeah. Yeah. Yeah. Understand.  What's coming for Hudl in terms   of the platform? Are there any neat things  that you're focused on, or even you're just   dreaming about building that you can talk about? Yeah. Absolutely. Again, two big things that I'm  

really excited about that we're working on. The  first, as just coming hot off our conversation   here with HEVC, we are looking to upgrade  our platform to provide HEVC for our coaches.  Amazing. We're really excited that we're going to   be able to deliver a visually lossless quality to  them, which is really big for us. The other one,   I would say, really excited about this technology,  is what we call our on-demand rendering system.   This is something that we are currently fully  utilizing for all of our secondary qualities   in the US. So all US-based accessed video is  currently using this in production, but it is  

the ability to encode the secondary renditions  of our ABR ladder, so our lower qualities,   on the fly. So this is something where we're  able to cut down our storage footprint in half,   which, when you're talking about the sheer  size and the amount of content we have-  Yeah. That's very meaningful. Yeah. When you think about it, in a lot   of situations I would say that the math wouldn't  work out and you'd be spending significantly more.   Our access patterns are that this coaching video  is generally going to be accessed over the first   14 days of its life. Then it will just sit there,  and the chances of it being accessed again are   very small. So if you can deliver those- What is your SLA? How long do you have to  

keep that video on the system? Yeah. We used to say it was   indefinite. It's going to be based on  the coach's subscription, their package-  I see. I see. ... how many hours   of video. We got to keep that video around as  long as the coach keeps it in their account. 

I see. I see. Okay. So is this, I  buy a bundle of hours, let's say,   and then I'm about to run out, and then I as a  coach have to decide, I really don't need those,   and I could, whatever, choose for deletion  or something, or just delete them, I guess-  Right. ... on video? Yeah. Got it. But yeah,   it's an interesting problem, and it's interesting  that you're pointing out that it's now becoming   possible to trade compute cycles, or I'm sorry,  trade storage for compute cycles, meaning that   it's actually more cost-effective to transcode on  the fly and not store those additional renditions,   than it is to just have them prerecorded  and stored. I've heard pretty consistently,   and of course, obviously we provide methods to  do that very, very cost effectively. So for a   lot of our customers, they've done that math and  it makes sense in almost every context to do it   that way. So yeah. Yeah, storage cost, even  though storage cost is being driven... it's   more benefited from Moore's law. Storage is still cheap, right? 

Yeah. Storage is cheap until you're talking  hundreds and hundreds and hundreds of petabytes.   And when you get up to a certain scale, there was- It's no longer cheap.  Yeah. Yeah. Yeah. It becomes meaningful. Yeah. Casey, I really want to thank you for the   conversation. It was wonderful. One of the things  that we work really hard to do on Voices of Video   is to have meaningful conversations just like  this one. You shared a lot of great insights  

and can't wait to have you back. Maybe after  you get HEVC deployed a little more fully,   we can have you come back and talk about  lessons learned and give us an update.  Yeah, and thanks for having me.  It was a great conversation.

2023-11-20 10:01

Show Video

Other news