WEBVTT
00:00:01.679 --> 00:00:04.799
Hello, and welcome to the OnRecord Podcast.
00:00:05.040 --> 00:00:16.800
This is a podcast where we have the opportunity of speaking with subject matter experts that are using video and audio in innovative ways to support their organization's objectives.
00:00:17.039 --> 00:00:28.399
In today's episode, we're taking a unique turn and we're going to explore some of the underlying technology that supports the Vault video platform.
00:00:28.640 --> 00:00:30.320
My name is Mike Anselone.
00:00:30.480 --> 00:00:32.320
My co-host is Carmen Hawk.
00:00:32.640 --> 00:00:34.880
And today, our guest is Barry Owen.
00:00:35.039 --> 00:00:38.719
Barry is the chief solution architect with WoWza.
00:00:39.200 --> 00:00:45.520
Wowza is the streaming engine that really drives the video component of Vault.
00:00:45.920 --> 00:00:47.679
So welcome to the podcast, Barry.
00:00:48.079 --> 00:00:48.399
Thanks.
00:00:48.560 --> 00:00:49.439
Nice to be here.
00:00:49.679 --> 00:00:59.759
So before we get started, can you tell us a little bit about yourself, uh, how you got um started with WoWza and what your you know overall mission is?
00:01:00.320 --> 00:01:00.880
Sure.
00:01:01.119 --> 00:01:07.359
Um I've always been one of those curious problem solver types of people.
00:01:07.519 --> 00:01:22.560
So, you know, I I went down the engineering route in school and became a programmer out of school and you know, did all kinds of engineering-y type things, but was always looking for, you know, new challenges and and things that changed.
00:01:22.640 --> 00:01:28.799
And I was I was drawn to video after a roundabout way of of many other things.
00:01:28.959 --> 00:01:32.079
I feel like I always had something to do with visuals and my products.
00:01:32.159 --> 00:01:35.120
I worked on a lot of image processing products and things like that.
00:01:35.359 --> 00:01:39.840
And you know, as video became more prevalent, it just was super interesting to me.
00:01:40.000 --> 00:01:48.719
And I had known the WoWSA founders from a past life at Adobe, and the timing just worked out for me to come join WoWSA.
00:01:49.200 --> 00:01:57.040
And at that point, I started and you know, started in the engineering organization running engineering, and then I've moved on to several other things.
00:01:57.120 --> 00:02:11.199
And now I have more of an outward-facing role as the chief solutions architect, which is really all about um understanding customers' needs and how video can help solve those.
00:02:11.439 --> 00:02:13.520
You know, and there's lots of challenges in that.
00:02:13.599 --> 00:02:20.639
You know, every everybody's streaming now, and streaming is in some in some extent, you know, somewhat mature.
00:02:20.960 --> 00:02:26.800
Um, used to be getting live video out at scale and securely and reliably was a huge challenge.
00:02:27.039 --> 00:02:30.479
And you know, that's reasonably well done at this point.
00:02:30.639 --> 00:02:32.879
And now we're finding new problems to solve, right?
00:02:32.960 --> 00:02:39.280
I mean, um how can you leverage that video to have real-time insights?
00:02:39.439 --> 00:02:50.479
How can you actually make sure that video is solving a problem and not just wind up being recorded and stashed in a drive somewhere in the cloud and never to be heard from again?
00:02:50.719 --> 00:02:57.599
Um, you know, lots of organizations are sitting on tons of video that you know nobody's watching.
00:02:57.759 --> 00:02:59.840
Um, and that's that's a missed opportunity.
00:03:00.000 --> 00:03:13.039
I mean, it could be in the live sense, it could be a missed opportunity to detect something that could save somebody's life or um, you know, detect something that could make some experience better for the user.
00:03:13.280 --> 00:03:37.280
So there's a lot of really interesting things happening now, and I think you know, you'll see a lot more uses of artificial intelligence within that stack to kind of help automate and solve some of those problems that are traditionally um some of the ways, some of the only ways you could do that was by throwing a lot of people at them.
00:03:38.159 --> 00:03:46.080
When you mentioned just now all that video content that people are sitting on, it reminds me of the 5th of July after all my fireworks videos.
00:03:46.400 --> 00:03:49.039
What am I gonna do with them the following year or the next year?
00:03:49.199 --> 00:03:52.159
But also we're seeing a lot of the alignment you're just describing.
00:03:52.400 --> 00:03:55.439
Why you are the chief solution architect for your organization?
00:03:55.599 --> 00:03:58.879
You are coming up with the solutions and you're being really creative.
00:03:59.039 --> 00:04:01.439
What is a current challenge right now?
00:04:01.599 --> 00:04:02.560
You're speaking about AI.
00:04:02.639 --> 00:04:11.680
What is a current challenge that your customers bring to you on a broad level or very niche that you would explore with us that you're helping to solve with AI?
00:04:12.319 --> 00:04:18.800
I'll give you an example that was from we just we just launched a lot of our AI capabilities this week.
00:04:18.879 --> 00:04:21.279
And we're, you know, we had a beta program leading up to that.
00:04:21.519 --> 00:04:25.040
We had some really interesting examples, and it's stuff you probably wouldn't think of, right?
00:04:25.279 --> 00:04:31.920
So um one of our customers is a is a company called Um Oceaneering.
00:04:32.399 --> 00:04:39.839
And what Oceaneering does is they do a lot of offshore monitoring of oil rigs and harsh environments and things like that.
00:04:40.160 --> 00:04:59.920
So areas where it's really not practical to do things like detect if a pipe is leaking or or you know, see if a pressure gauge on some system is too high, or even things like motion on the rig itself, you know, if someone's using a crane to tell people to stay away from that area.
00:05:00.160 --> 00:05:15.759
But some of the cool things they were able to do is they were able to, you know, via a very remote camera under the ocean, you know, see a little tiny stream of bubbles coming out of a pipe and recognize that that's actually bubbles coming out of a pipe and not something else.
00:05:15.920 --> 00:05:23.199
And that's applying this whole system to really understand it's like, hey, this could be an issue.
00:05:23.519 --> 00:05:32.079
We can actually send that alert off in real time so somebody can determine, you know, then somebody because again, how many of those things do they actually have to watch?
00:05:32.240 --> 00:05:34.240
Nobody's watching all that video in real time.
00:05:34.319 --> 00:05:35.519
It's impossible.
00:05:35.759 --> 00:05:44.800
Um so we can trigger an alert and allow people to draw attention to something that needs to be drawn attention to, and they can then take the appropriate action.
00:05:44.959 --> 00:05:46.959
And I think that's really interesting use case.
00:05:47.040 --> 00:05:51.839
And and you can generalize that into you know, tons of other stuff, right?
00:05:51.920 --> 00:06:06.399
I mean, the the vault platform could apply AI to, you know, detect a clinical study where, I don't know, someone's not in bed when they're supposed to be in bed, or someone's not nursing the position they're supposed to be in.
00:06:06.560 --> 00:06:11.120
Um, there's all kinds of interesting use cases you can you can apply to this.
00:06:12.480 --> 00:06:16.879
So, Barry, WoWSA, first of all, I think is a really cool name for a company.
00:06:17.199 --> 00:06:23.439
Um can you define what the WoWSA streaming engine is and how it works?
00:06:24.399 --> 00:06:29.920
Yeah, I usually I usually tell people what you know, Wowza provides streaming infrastructure for people.
00:06:30.079 --> 00:06:36.480
So WoWza streaming engine is this very sophisticated kind of end-to-end platform.
00:06:36.639 --> 00:06:45.680
So we can take video in from a variety of sources, whether it be a camera, a file, um, a satellite link, whatever it might be.
00:06:45.920 --> 00:06:54.720
And we can take that video and we can we can process it, we can package it for delivery to any device anywhere in near real time.
00:06:55.120 --> 00:06:58.319
And then we can also do analysis on it along the way.
00:06:58.399 --> 00:07:06.879
Um, we can make sure that video is secure, we can encrypt it if it needs to be encrypted, we can make sure only the people who are supposed to be watching the video are supposed to be watching the video.
00:07:07.199 --> 00:07:17.519
And it's anywhere from, you know, hundreds of inputs to very few people watching, or machines watching, or single inputs to millions of people watching, right?
00:07:17.680 --> 00:07:18.959
It scales both ways.
00:07:19.040 --> 00:07:24.560
And Streaming Engine was built to be this kind of general purpose workhorse in the middle of all that.
00:07:24.800 --> 00:07:34.079
Um it's very API driven, so it's often built into other products, like the folks at IVS have used with Vault.
00:07:34.240 --> 00:07:35.759
Is the engine is built in.
00:07:35.920 --> 00:07:38.399
You know, it's part of the, it's part of the product.
00:07:38.560 --> 00:07:41.439
It's not necessarily used by itself.
00:07:41.519 --> 00:07:47.120
It's it's part of the offering and it's built into the to the product at a very deep level.
00:07:48.079 --> 00:07:53.759
And one of the questions a lot of our clients ask is how scalable is Vault?
00:07:54.639 --> 00:08:01.120
And that scalability really is built on, in large part, the WoWSA streaming engine.
00:08:01.279 --> 00:08:07.439
Can you talk a little bit about the topic of scalability and how you make that possible?
00:08:08.240 --> 00:08:08.800
Sure.
00:08:09.040 --> 00:08:11.360
Um, a couple different ways, right?
00:08:11.519 --> 00:08:19.759
So streaming engine itself is built to take maximum advantage of any underlying hardware that it can.
00:08:20.319 --> 00:08:29.199
So things like video processing lend well to being processed with GPUs or VPUs to really increase the density of the video.
00:08:29.439 --> 00:08:50.960
So if you have a thousand cameras you need to process, um, depending on how you want to deploy your architecture, you can take advantage of those to make highly efficient um workflows and highly efficient hardware setups without having to have tens or hundreds of servers necessarily running.
00:08:51.200 --> 00:08:55.440
Um, I often tell people it's also a deployment flexibility, right?
00:08:55.519 --> 00:09:03.919
So streaming engine can run on-prem, it can run on little edge devices, it can run in the cloud or a combination of all of them.
00:09:04.159 --> 00:09:09.039
So I often tell people it's like, you know, what's your most important aspect?
00:09:09.279 --> 00:09:11.600
Are you looking to scale up and down a lot?
00:09:11.759 --> 00:09:18.639
You know, so we have customers who maybe have a hundred streams going all the time, but maybe on the weekends they might go up to a thousand streams.
00:09:18.960 --> 00:09:30.799
So I would probably architect their solution a little differently so it can scale up and down very quickly and efficiently, versus someone that maybe they have 10 streams that are going out to lots and lots of viewers.
00:09:31.039 --> 00:09:33.759
Maybe they need some very fixed infrastructure.
00:09:34.000 --> 00:09:47.759
So it's really dynamic and it's really dependent on customer use cases, which is, you know, honestly kind of one of the fun parts about my job is I get to see all kinds of different uses of this and often used in ways you wouldn't expect.
00:09:48.559 --> 00:09:51.120
What are some you mentioned one use case?
00:09:51.360 --> 00:09:55.919
What are some of the more interesting use cases that you've come across?
00:09:56.639 --> 00:09:57.759
Oh, tons.
00:09:58.000 --> 00:10:07.759
So um we stream live video from Formula One cars to the pits, um, both with telemetry and track data and things like that.
00:10:08.000 --> 00:10:11.679
Um, we have streamed live video from the moon.
00:10:12.080 --> 00:10:18.879
So being part of the NASA project to stream live video from Artemis was an amazing experience.
00:10:19.279 --> 00:10:22.480
So, you know, being part of things like that are great.
00:10:22.639 --> 00:10:40.480
And then, you know, products like products like yours, products like um, you know, we have customers who are doing e911 services so that operators can, you know, get real-time video information um from people who are calling 911.
00:10:40.799 --> 00:10:44.000
So it really does range all over the place.
00:10:44.480 --> 00:10:47.759
And you know, there's always something very interesting.
00:10:48.559 --> 00:10:54.320
You've mentioned some savings that was um related to the oil rig that's monitoring underwater.
00:10:54.559 --> 00:11:06.240
Um you've also talked about like the scalability of the solution and the stability that's required of a solution that's going to be supporting moon live capture, um, formula one to pit capture.
00:11:06.399 --> 00:11:11.279
Um, what are some other types of problem solving that your company is doing?
00:11:11.440 --> 00:11:17.519
We see a lot of overlap with those sentences I just uttered, the reliability, scalability, cost effectiveness in our solution.
00:11:17.679 --> 00:11:28.960
Um, just curious if there's anything else from a from a uh presentation perspective that's helpful for prospects to know that um your team does exceptionally well.
00:11:30.720 --> 00:11:40.639
Well, so with the new tooling and all the the AI capabilities, what we're really talking about is trying learning to treat video as a data source, right?
00:11:40.720 --> 00:11:48.799
So, how can we treat video as data and really make that data in that video as effective as possible for actions?
00:11:49.679 --> 00:11:58.240
So making video an actionable resource where perhaps it was more of a passive or non-real time before.
00:11:58.480 --> 00:12:07.840
Um, I would say our team has been really, you know, on the engineering and product side, we've been really leaning into that to look at really interesting use cases for that.
00:12:08.000 --> 00:12:22.080
On our implementation and support side, it's it's really that feedback from the customers, how can these little things we're doing add up into big changes for a customer's outcome?
00:12:22.559 --> 00:12:43.600
And I think you know, the overarching scheme of that is really how can we how can we turn this into a first class data source where um perhaps in the past it was you know it was just missing that component.
00:12:44.399 --> 00:12:45.360
It's interesting.
00:12:45.600 --> 00:12:58.799
We um, you know, we spend a lot of time in the healthcare market working with like nursing schools, medical schools, um, where they're doing simulation.
00:12:58.960 --> 00:13:19.600
And one of the big drivers in that marketplace is collecting video as data to support an evidence-based framework, so that evidence being able to really um show learner competency in a particular field of discipline.
00:13:20.000 --> 00:13:32.320
So I'm curious from from an analytics perspective, what how WoWza may be looking at that kind of application as well.
00:13:32.799 --> 00:13:38.559
I think as long as you can put heuristics around what you're looking for, right?
00:13:38.799 --> 00:13:45.679
So, what determines that competency and what does that look like in a video scenario?
00:13:46.320 --> 00:13:58.320
Um, I think then you could probably, you know, we don't pretend to have every model needed for everybody's video.
00:13:58.799 --> 00:14:08.799
We designed an open platform for people to be able to add models or train models or tune models to really meet their precise needs.
00:14:09.039 --> 00:14:16.639
So I can envision somebody coming up with um something that's tuned to those clinical settings.
00:14:16.879 --> 00:14:25.360
And maybe it's okay, if you notice that A happened within 10 seconds, you should be expecting B to happen.
00:14:25.759 --> 00:14:33.679
And if you don't see B, maybe you do some sort of alert or some sort of a notification, or even something as simple as marking the time in the video.
00:14:34.240 --> 00:14:46.799
And then you start to build that over time, and you can take that data back in your models and keep training them for even, you know, more input and more correctness, if you will.
00:14:46.960 --> 00:14:57.279
Um, so really it's gonna be about the ability to again see into the video and and notice things like that.
00:14:58.159 --> 00:15:01.759
So I have a question when it's related to healthcare.
00:15:01.840 --> 00:15:04.720
Just you mentioned a healthcare use case, as Mike also alluded to.
00:15:04.960 --> 00:15:06.639
Healthcare is an industry we both serve.
00:15:06.799 --> 00:15:08.720
We have law enforcement research.
00:15:09.200 --> 00:15:11.039
These are really sensitive spaces.
00:15:11.279 --> 00:15:17.039
Um, if they're interested in using video AI, what is different about how we serve as partners?
00:15:17.120 --> 00:15:22.720
Um, or how WoW's a service as partners when they're looking to use video AI in those sensitive industries?
00:15:23.120 --> 00:15:29.440
Yeah, I think there's a couple things, is we're not sending the video off to the cloud to be analyzed.
00:15:29.759 --> 00:15:36.799
So with Streaming Engine and the VIP product, which is the video intelligence framework, it all runs on-prem.
00:15:36.960 --> 00:15:40.559
It can run on the same devices you're running your videos now.
00:15:40.720 --> 00:15:43.840
You know, you don't have to go in and rip and replace a bunch of stuff.
00:15:44.000 --> 00:15:50.799
You're not sending video off to a cloud service where you lose control of it, where it gets parked in a cloud environment somewhere.
00:15:51.039 --> 00:15:56.080
This can all happen, you know, in real time, on the edge or on-prem.
00:15:56.399 --> 00:16:00.240
And you know, Streaming Engine was meant to be deployed in those kind of environments, right?
00:16:00.320 --> 00:16:05.279
It's it gets built in, it can be deployed in a hospital within the hospital's network.
00:16:05.360 --> 00:16:26.559
It can deploy in, you know, we actually are deployed in police stations and places like that where, you know, not to mention government and other secure installations, where you know, those are private networks, you architect solutions such that you have the resources necessary to handle the scale they need to handle at.
00:16:26.639 --> 00:16:38.480
And that could be anywhere from a single small edge machine or a laptop on up to multiple servers and then you know, multiple GPUs to handle both the AI and the video processing.
00:16:38.720 --> 00:16:47.759
So it's it's really nice that you don't have to leverage a cloud service or a SaaS platform to get this.
00:16:48.480 --> 00:16:57.360
I know AI is everybody that we're talking to, that's usually their number one question is how are you using AI?
00:16:57.519 --> 00:17:02.159
And I'm kind of curious because it it's evolving so rapidly.
00:17:03.360 --> 00:17:03.840
For sure.
00:17:04.400 --> 00:17:12.640
What's Wows's roadmap like a year to even five years down the road in implementing AI in your technology?
00:17:13.039 --> 00:17:19.119
I think there's a few things, and you're you're starting to see an inkling of what some of those can be with the product re-release.
00:17:19.200 --> 00:17:23.200
So, you know, we started with relatively straightforward object detection.
00:17:23.519 --> 00:17:26.880
So, you know, hey, look at a video, there's cars there.
00:17:27.039 --> 00:17:29.920
Great, you can count the cars, whatever it is.
00:17:30.079 --> 00:17:37.599
Um, then you move to things like, okay, now I see cars going both directions on a highway.
00:17:37.759 --> 00:17:39.039
Can I count them?
00:17:39.359 --> 00:17:42.880
Now can I tell you if the cars are going the right direction or not?
00:17:43.200 --> 00:17:48.880
And then you're starting to get into even more dynamic things like, okay, now I see a bunch of cars on a highway.
00:17:49.839 --> 00:17:58.000
Can I predict if something's going to happen, whether it be a traffic jam or an accident or something like that, based on the scenes of data I'm seeing?
00:17:58.559 --> 00:18:09.119
And then you can you can move into the reason we designed the framework like we did is because we want to be able to easily add these new models and these new capabilities as they evolve.
00:18:09.359 --> 00:18:15.839
So one of the models we did work with was a model from NVIDIA that is all about synthetic video detection.
00:18:16.480 --> 00:18:21.440
So what you're starting to see is obviously more and more and more AI generated video.
00:18:22.160 --> 00:18:31.039
And from a perspective of you know news, social media, whatever it all, you might want to be able to detect that.
00:18:31.279 --> 00:18:36.400
That doesn't necessarily mean that synthetic generated video is bad.
00:18:36.640 --> 00:18:41.680
It's just you want to be able to know if it's synthetic or if it's real and take the appropriate action.
00:18:41.839 --> 00:18:43.599
You know, a social media site may not care.
00:18:43.680 --> 00:18:49.680
It's like they may just label a day, I hey, this is you know, an AI generated video reading something.
00:18:50.000 --> 00:18:55.039
Um, but a news, you know, a news, a news source might not want that.
00:18:55.200 --> 00:19:01.200
They might want to make sure that they see if it's synthetic video, they do something completely different.
00:19:01.359 --> 00:19:03.519
So things like that will keep evolving.
00:19:03.759 --> 00:19:13.680
And then you'll be able, as the models get more powerful, and you'll be able to do more things across scenes of video, right?
00:19:13.759 --> 00:19:19.759
You can already do things like context detection, and that's good for a bunch of different reasons.
00:19:19.839 --> 00:19:31.519
But what it's what it's starting to be more commonly used for is making sure that an upcoming advertisement in a video stream isn't wildly different than the content before it.
00:19:31.599 --> 00:19:38.960
You know, you don't want a happy ad with a sad, you know, you see sad scene in the movie, all of a sudden you see puppies.
00:19:39.200 --> 00:19:43.759
Um you do, but but you know, you you at least want to know.
00:19:44.000 --> 00:19:46.319
Um so things like that.
00:19:46.960 --> 00:19:49.839
You're starting to see this go a lot of different directions.
00:19:49.920 --> 00:20:01.039
But I think some of the early wins are certainly the kind of chaining of events together to determine outcomes or if an outcome may happen and some predictions.
00:20:01.519 --> 00:20:10.480
And also, you know, the there's huge problems with content moderation across user-generated content services and social media.
00:20:10.799 --> 00:20:13.920
So better uh better applying AI to that.
00:20:14.160 --> 00:20:18.799
Um so you don't have to just solve the problem with you know teams of people.
00:20:19.599 --> 00:20:24.240
Uh, because to be honest, you know, nobody really wants to watch that video in the first place.
00:20:24.400 --> 00:20:33.039
And certainly by just teams of teams of people, you're not only subjecting them to having to watch that video, but they they miss stuff.
00:20:33.440 --> 00:20:36.640
I like everything you just mentioned from a from an ethical perspective.
00:20:36.720 --> 00:20:46.480
It sounds very um where wherever you sit on the spectrum of your AI approval, everything that um you just mentioned is all like with good intent and to do good.
00:20:46.640 --> 00:20:49.119
So we hope, right?
00:20:49.359 --> 00:20:49.839
Yeah.
00:20:50.960 --> 00:20:54.960
Hey Barry, what makes the Wowza streaming engine unique?
00:20:56.319 --> 00:21:02.880
I think, well, so Streaming Engine is is kind of the OG of streaming engines, right?
00:21:02.960 --> 00:21:05.440
We launched our first product in 2007.
00:21:06.000 --> 00:21:14.960
And you know, from the ground up, it was built to be customized and added to by users, right?
00:21:15.039 --> 00:21:22.799
So it's got a full API for you to not only for command and control, but also for extensibility.
00:21:23.200 --> 00:21:31.599
So earlier this year, as an example, um, I decided I wanted to learn how to use Cloud to write code.
00:21:31.680 --> 00:21:35.200
Yeah, I wrote code all my life, but I haven't written a lot of code in the last 10 years.
00:21:35.359 --> 00:21:38.079
So I was like, well, I'll see what Cloud can help me do.
00:21:38.319 --> 00:21:52.079
And I was able to add um a new streaming protocol to Streaming Engine over the course of really just a couple of weeks using engine APIs only.
00:21:52.160 --> 00:21:59.759
I didn't have to touch any engine code, but I was able to add the protocols called mock, and it's kind of an up and coming streaming protocol for.
00:22:00.079 --> 00:22:06.720
Low latency, potentially to replace WebRTC, but uh potentially replace a bunch of a bunch of other protocols.
00:22:06.960 --> 00:22:33.839
Um but the fact that I was able to do that without having to change any code in engine itself gives you a glimpse into the other how powerful those APIs are that you could truly integrate engine very deeply into whatever workflow you need to integrate it into without having to you know worry about backward compatibility or or anything changing.
00:22:33.920 --> 00:22:36.880
And you know, you can control that code moving forward.
00:22:37.680 --> 00:22:42.000
You mentioned Web RTC, I'm which kind of piqued my curiosity.
00:22:42.480 --> 00:22:46.960
How is mock different than Web RTC?
00:22:47.920 --> 00:22:49.680
We might not have enough time for this.
00:22:49.839 --> 00:22:53.200
Uh we can use claude to summarize.
00:22:54.319 --> 00:23:14.400
Mock is a a new streaming protocol kind of designed from the ground up to take to take advantage of some newer protocols like what's called QIC, which is part of HTTP3, which is basically a UDP-based protocol that can be used in the browser.
00:23:14.799 --> 00:23:17.839
And there's significant advantages to that for streaming.
00:23:18.079 --> 00:23:25.119
TCP has a lot of overhead, um, and it builds latency when there's when there's packet loss because it does retransmissions.
00:23:25.680 --> 00:23:35.359
And you know, traditional streaming right now, we have you know, we have HLS and Dash, which are the most common things for one-to-many streaming, right?
00:23:35.599 --> 00:23:41.599
If you're watching Netflix, if you're watching Apple TV, whatever it is, it's going to be HLS or Dash.
00:23:41.920 --> 00:23:47.200
Um, WebRTC was originally designed for video conferencing, still does a pretty good job of that.
00:23:47.359 --> 00:23:49.200
It's not as great at one-to-many.
00:23:49.359 --> 00:23:51.839
Um, it has problems in other areas.
00:23:52.079 --> 00:23:57.519
It wasn't really designed for things like accessibility, and it wasn't designed for things like DRM.
00:23:58.000 --> 00:24:10.960
So basically what Media Overquick is is attempting to do is take a new look at streaming based on the technology we know now, rather than evolved over the last 10 or 15 years.
00:24:11.200 --> 00:24:13.759
It's like, how would we do this if we were starting over?
00:24:14.480 --> 00:24:16.960
And kind of trying to be the best of all worlds.
00:24:17.039 --> 00:24:18.319
And it's got a lot of cool features.
00:24:18.480 --> 00:24:20.880
You can do super low latency with it.
00:24:21.039 --> 00:24:23.839
Um, you know, it's it's playable in browsers.
00:24:23.920 --> 00:24:28.559
It's you can broadcast from a browser, but you can also broadcast from a camera.
00:24:28.880 --> 00:24:31.039
Um, it scales really well.
00:24:31.200 --> 00:24:34.640
You can kind of tune the latency for your audience size.
00:24:35.039 --> 00:24:45.119
It's designed to be deployed with what we call relays, which you send a stream to a relay, that relay can then send streams to people or other relays.
00:24:45.279 --> 00:24:48.000
So you get this kind of treed fan out effect.
00:24:48.319 --> 00:24:50.240
So there's a lot of promise there.
00:24:50.319 --> 00:24:51.839
Um it's getting close.
00:24:52.160 --> 00:24:56.559
You can certainly go um read lots about it.
00:24:56.720 --> 00:25:05.920
Open mock is an uh industry group that's kind of deploying versions of it as the spec evolves.
00:25:06.160 --> 00:25:15.920
Um, but I think it's going to be a really interesting, interesting alternative to some of the other um protocols over the next couple years.
00:25:16.319 --> 00:25:33.759
What kind of guidance and recommendations can you offer organizations that may either be looking to upgrade some of their video streaming capabilities or just getting started out in terms of, yep, we recognize a need uh to stream video.
00:25:34.000 --> 00:25:41.440
It could be a nursing simulation that wants to stream from a live simulation to a classroom of students, for example.
00:25:41.599 --> 00:25:46.319
Um whether that classroom of students is across the hall or across the country.
00:25:46.640 --> 00:25:47.039
Yep.
00:25:47.359 --> 00:25:54.480
And they're kind of getting started, like, well, you know, the big question could be where and how do we get started?
00:25:55.759 --> 00:26:00.960
I my first advice would would, you know, start small, right?
00:26:01.200 --> 00:26:09.759
So pick one clear use case, one workflow, one problem that you're trying to solve, and get it working end to end, and get it working well.
00:26:10.240 --> 00:26:11.119
We'll take your case.
00:26:11.200 --> 00:26:19.119
You know, I have I have one class I'm trying to stream from one place to another.
00:26:19.759 --> 00:26:30.400
Get that working, get the recording working, get the interactivity elements working, get that dialed in, and then you will you'll understand what you need to do that at scale.
00:26:31.200 --> 00:26:41.680
So, you know, I would say don't try to jump in and do 200 classes next week to you know 200 different sites.
00:26:42.160 --> 00:26:44.000
Um get it working.
00:26:44.319 --> 00:26:50.079
And then many times you have the infrastructure already.
00:26:50.240 --> 00:26:52.400
You have cameras, you may have servers.
00:26:52.480 --> 00:26:58.480
You don't necessarily need to go out and buy a whole broadcast studio's worth of infrastructure to get started.
00:26:58.720 --> 00:27:03.920
You know, start with what you have because you know it may be more than good enough to get you there.
00:27:04.240 --> 00:27:09.599
Um and then, you know, find people to help you, right?
00:27:09.759 --> 00:27:23.279
I mean, whether it be vendors or whether it be communities, but there's a lot of there's a lot of places to get advice on you know representative best practices and how to do things.
00:27:23.440 --> 00:27:30.720
I'm not I'm not gonna say the right way to do things because that's different for every implementation and different for every person's use case.
00:27:30.799 --> 00:27:36.799
But there's a lot of yeah, there's a lot of resources out there for you to take advantage of.
00:27:37.039 --> 00:27:40.160
And that's where I think you know, Wowza certainly is great.
00:27:40.240 --> 00:27:43.359
We have a great support team, we have a great services team.
00:27:43.519 --> 00:27:48.720
They understand not only Wowza's products, but they understand a lot of other products and how things all work together.
00:27:48.799 --> 00:27:50.400
And I think that's important too.
00:27:50.720 --> 00:27:52.480
So you mentioned infrastructure.
00:27:52.880 --> 00:28:01.279
I'm kind of curious what kind of infrastructure requirements are you looking for in order to kind of deploy your solutions?
00:28:02.640 --> 00:28:04.640
So we have a ton of flexibility in that regard.
00:28:05.359 --> 00:28:12.400
As I mentioned, we can deploy on tiny little edge servers, you know, not much bigger than a Raspberry Pi, to be honest.
00:28:12.480 --> 00:28:14.960
They're very, very small little devices.
00:28:15.200 --> 00:28:21.440
And if you only have a camera or two, you're looking to get repackaged and streamed out for mobile consumption or whatever it is.
00:28:21.680 --> 00:28:22.880
That may be enough.
00:28:23.279 --> 00:28:29.440
And if you have a remote setting where you're trying to do something, maybe plunking one of those down in that remote setting is gonna be good.
00:28:29.920 --> 00:28:34.319
And you know, you have options as far as what your input source might be.
00:28:34.400 --> 00:28:46.160
You could you could be streaming from a phone into that box, you could be streaming from a you know a dedicated camera or even a professional camera, and you know, you can start there.
00:28:46.400 --> 00:28:53.920
And you know, typically people are these days is almost most streams are going to be viewed um on a mobile device.
00:28:54.640 --> 00:29:01.279
So optimize your stream so it works well on a mobile device, it works well under all different bandwidth conditions.
00:29:01.440 --> 00:29:11.200
You know, and streaming engine is great at creating sets of outputs that look good on mobile devices under varying bandwidth conditions and varying screen sizes and things like that.
00:29:11.519 --> 00:29:12.720
I like that advice you gave.
00:29:12.799 --> 00:29:17.759
And I'm curious what type of uh internal organizational context do you work with?
00:29:18.079 --> 00:29:25.119
Do you work with a lot of end users or do you work with the network team or who are you really collaborating with from the ground up?
00:29:25.519 --> 00:29:27.839
So it's it's kind of a combination, right?
00:29:28.079 --> 00:29:38.480
So I will typically talk to the business owners of the problem to understand what it is they're trying to accomplish.
00:29:38.799 --> 00:30:02.559
But then when it comes to architecture and implementation, it's it's someone in the technical team, whether whether they have a video engineering group or whether it's the IT team or a combination of the two, um, will work with them to understand their infrastructure constraints or opportunities on the deployment side and how to set up a solution.
00:30:02.640 --> 00:30:07.440
And that's where we often also will say, you know, bring in other vendors.
00:30:07.599 --> 00:30:11.759
You know, you may need a delivery provider to deliver things at scale.
00:30:12.079 --> 00:30:18.160
You may need, you may need a cloud provider at some point to step in to do massive storage or something.
00:30:18.400 --> 00:30:27.039
So understanding where things are and what things needs to happen will also influence um the kinds of partners you might need to bring into the solution.
00:30:27.359 --> 00:30:32.240
And you help with that overall design so you can help make rough like referrals or you can help uh guide.
00:30:34.160 --> 00:30:40.079
We we have tons of partners in the industry, not only from like a systems integration perspective, but vendors.
00:30:40.160 --> 00:30:46.319
Like if we if we need a DRM vendor, we can bring in a DRM vendor we've worked with in the past that we know is going to work.
00:30:46.559 --> 00:30:50.319
Um, encoders, cameras, you know, CDNs.
00:30:50.880 --> 00:31:01.680
Um so we definitely have the whole aspects of the ecosystem covered as far as how we attack problems.
00:31:03.839 --> 00:31:12.880
So, Bear, before we wrap up, is there anything else that you'd like to share with our audience or questions that we should have asked that we missed?
00:31:15.599 --> 00:31:16.480
Let's see.
00:31:16.799 --> 00:31:25.440
Um I think let me look through my notes.
00:31:25.680 --> 00:31:31.519
Um I think, I mean, we we covered a wide range of ground, and I think it's interesting.
00:31:31.599 --> 00:31:54.000
And I think, you know, what I would what I would you know say if you're streaming now, um take a look how you're actually using that video, and if some of the new things coming out in terms of of AI and and video intelligence might help you.
00:31:54.240 --> 00:31:57.920
Um the answer may be not right now, and that's fine.
00:31:58.160 --> 00:32:20.240
But you know, I think what you'll find is that there's data buried in that video, and you're really looking to make the most out of your infrastructure and your video investments, building that case for, hey, we've done great with video now, but look at all this stuff that's actually opening up and we can we can look at and do some experiments, right?
00:32:20.319 --> 00:32:24.079
I mean, it's um whatever it might be.
00:32:25.599 --> 00:32:27.519
Well, that's awesome advice.
00:32:28.079 --> 00:32:34.480
How can our listeners get in touch with you if they they want to have a deeper conversation?
00:32:35.839 --> 00:32:37.759
Well, they can certainly find me on LinkedIn.
00:32:38.000 --> 00:32:39.359
Pretty easy to find there.
00:32:39.519 --> 00:32:40.880
It's Barry J.
00:32:41.039 --> 00:32:43.519
Owen, whatever the whole LinkedIn thing Barry J.
00:32:43.680 --> 00:32:44.559
Owen is.
00:32:44.960 --> 00:32:47.759
Um, you can find me on waza.com.
00:32:47.839 --> 00:32:49.839
I write a lot of blog posts there.
00:32:50.480 --> 00:32:59.359
And you know, occasionally you will see me speaking at events or at conferences, certainly at trade shows.
00:32:59.519 --> 00:33:01.839
Um, but I'd love to hear from people.
00:33:02.079 --> 00:33:04.799
Happy to connect on LinkedIn and talk more.
00:33:05.279 --> 00:33:09.519
And you know, I'm not going to come at you with a hard sell.
00:33:10.000 --> 00:33:13.279
I, you know, you must buy WoWza because of this.
00:33:13.440 --> 00:33:22.160
Um, I really want to come at you more of what problem are you trying to solve and what's the right way to go about it and how you know how are you going to get the most out of what you're doing?
00:33:22.480 --> 00:33:23.680
What a great resource.
00:33:23.920 --> 00:33:24.240
Yeah.
00:33:24.480 --> 00:33:25.759
Well, thanks for your time, Barry.
00:33:25.839 --> 00:33:26.880
We really appreciate it.
00:33:26.960 --> 00:33:32.640
Um, I think our listeners have uh will learn a lot uh from our our podcast.
00:33:33.200 --> 00:33:33.920
Well, I hope so.
00:33:34.079 --> 00:33:34.559
It's fun.
00:33:34.720 --> 00:33:35.680
Appreciate the time.
00:33:35.839 --> 00:33:36.720
Thank you.
00:33:38.559 --> 00:33:46.160
OnRecord is a podcast presented by Intelligent Video Solutions, hosted by Mike Anzolone and Carmen Hawk, and produced by Kyle Shellstead.
00:33:46.319 --> 00:33:53.440
Find us online at ipivs.com slash on record and on social media at Intelligent Video Solutions.
00:00:01.679 --> 00:00:04.799
Hello, and welcome to the OnRecord Podcast.
00:00:05.040 --> 00:00:16.800
This is a podcast where we have the opportunity of speaking with subject matter experts that are using video and audio in innovative ways to support their organization's objectives.
00:00:17.039 --> 00:00:28.399
In today's episode, we're taking a unique turn and we're going to explore some of the underlying technology that supports the Vault video platform.
00:00:28.640 --> 00:00:30.320
My name is Mike Anselone.
00:00:30.480 --> 00:00:32.320
My co-host is Carmen Hawk.
00:00:32.640 --> 00:00:34.880
And today, our guest is Barry Owen.
00:00:35.039 --> 00:00:38.719
Barry is the chief solution architect with WoWza.
00:00:39.200 --> 00:00:45.520
Wowza is the streaming engine that really drives the video component of Vault.
00:00:45.920 --> 00:00:47.679
So welcome to the podcast, Barry.
00:00:48.079 --> 00:00:48.399
Thanks.
00:00:48.560 --> 00:00:49.439
Nice to be here.
00:00:49.679 --> 00:00:59.759
So before we get started, can you tell us a little bit about yourself, uh, how you got um started with WoWza and what your you know overall mission is?
00:01:00.320 --> 00:01:00.880
Sure.
00:01:01.119 --> 00:01:07.359
Um I've always been one of those curious problem solver types of people.
00:01:07.519 --> 00:01:22.560
So, you know, I I went down the engineering route in school and became a programmer out of school and you know, did all kinds of engineering-y type things, but was always looking for, you know, new challenges and and things that changed.
00:01:22.640 --> 00:01:28.799
And I was I was drawn to video after a roundabout way of of many other things.
00:01:28.959 --> 00:01:32.079
I feel like I always had something to do with visuals and my products.
00:01:32.159 --> 00:01:35.120
I worked on a lot of image processing products and things like that.
00:01:35.359 --> 00:01:39.840
And you know, as video became more prevalent, it just was super interesting to me.
00:01:40.000 --> 00:01:48.719
And I had known the WoWSA founders from a past life at Adobe, and the timing just worked out for me to come join WoWSA.
00:01:49.200 --> 00:01:57.040
And at that point, I started and you know, started in the engineering organization running engineering, and then I've moved on to several other things.
00:01:57.120 --> 00:02:11.199
And now I have more of an outward-facing role as the chief solutions architect, which is really all about um understanding customers' needs and how video can help solve those.
00:02:11.439 --> 00:02:13.520
You know, and there's lots of challenges in that.
00:02:13.599 --> 00:02:20.639
You know, every everybody's streaming now, and streaming is in some in some extent, you know, somewhat mature.
00:02:20.960 --> 00:02:26.800
Um, used to be getting live video out at scale and securely and reliably was a huge challenge.
00:02:27.039 --> 00:02:30.479
And you know, that's reasonably well done at this point.
00:02:30.639 --> 00:02:32.879
And now we're finding new problems to solve, right?
00:02:32.960 --> 00:02:39.280
I mean, um how can you leverage that video to have real-time insights?
00:02:39.439 --> 00:02:50.479
How can you actually make sure that video is solving a problem and not just wind up being recorded and stashed in a drive somewhere in the cloud and never to be heard from again?
00:02:50.719 --> 00:02:57.599
Um, you know, lots of organizations are sitting on tons of video that you know nobody's watching.
00:02:57.759 --> 00:02:59.840
Um, and that's that's a missed opportunity.
00:03:00.000 --> 00:03:13.039
I mean, it could be in the live sense, it could be a missed opportunity to detect something that could save somebody's life or um, you know, detect something that could make some experience better for the user.
00:03:13.280 --> 00:03:37.280
So there's a lot of really interesting things happening now, and I think you know, you'll see a lot more uses of artificial intelligence within that stack to kind of help automate and solve some of those problems that are traditionally um some of the ways, some of the only ways you could do that was by throwing a lot of people at them.
00:03:38.159 --> 00:03:46.080
When you mentioned just now all that video content that people are sitting on, it reminds me of the 5th of July after all my fireworks videos.
00:03:46.400 --> 00:03:49.039
What am I gonna do with them the following year or the next year?
00:03:49.199 --> 00:03:52.159
But also we're seeing a lot of the alignment you're just describing.
00:03:52.400 --> 00:03:55.439
Why you are the chief solution architect for your organization?
00:03:55.599 --> 00:03:58.879
You are coming up with the solutions and you're being really creative.
00:03:59.039 --> 00:04:01.439
What is a current challenge right now?
00:04:01.599 --> 00:04:02.560
You're speaking about AI.
00:04:02.639 --> 00:04:11.680
What is a current challenge that your customers bring to you on a broad level or very niche that you would explore with us that you're helping to solve with AI?
00:04:12.319 --> 00:04:18.800
I'll give you an example that was from we just we just launched a lot of our AI capabilities this week.
00:04:18.879 --> 00:04:21.279
And we're, you know, we had a beta program leading up to that.
00:04:21.519 --> 00:04:25.040
We had some really interesting examples, and it's stuff you probably wouldn't think of, right?
00:04:25.279 --> 00:04:31.920
So um one of our customers is a is a company called Um Oceaneering.
00:04:32.399 --> 00:04:39.839
And what Oceaneering does is they do a lot of offshore monitoring of oil rigs and harsh environments and things like that.
00:04:40.160 --> 00:04:59.920
So areas where it's really not practical to do things like detect if a pipe is leaking or or you know, see if a pressure gauge on some system is too high, or even things like motion on the rig itself, you know, if someone's using a crane to tell people to stay away from that area.
00:05:00.160 --> 00:05:15.759
But some of the cool things they were able to do is they were able to, you know, via a very remote camera under the ocean, you know, see a little tiny stream of bubbles coming out of a pipe and recognize that that's actually bubbles coming out of a pipe and not something else.
00:05:15.920 --> 00:05:23.199
And that's applying this whole system to really understand it's like, hey, this could be an issue.
00:05:23.519 --> 00:05:32.079
We can actually send that alert off in real time so somebody can determine, you know, then somebody because again, how many of those things do they actually have to watch?
00:05:32.240 --> 00:05:34.240
Nobody's watching all that video in real time.
00:05:34.319 --> 00:05:35.519
It's impossible.
00:05:35.759 --> 00:05:44.800
Um so we can trigger an alert and allow people to draw attention to something that needs to be drawn attention to, and they can then take the appropriate action.
00:05:44.959 --> 00:05:46.959
And I think that's really interesting use case.
00:05:47.040 --> 00:05:51.839
And and you can generalize that into you know, tons of other stuff, right?
00:05:51.920 --> 00:06:06.399
I mean, the the vault platform could apply AI to, you know, detect a clinical study where, I don't know, someone's not in bed when they're supposed to be in bed, or someone's not nursing the position they're supposed to be in.
00:06:06.560 --> 00:06:11.120
Um, there's all kinds of interesting use cases you can you can apply to this.
00:06:12.480 --> 00:06:16.879
So, Barry, WoWSA, first of all, I think is a really cool name for a company.
00:06:17.199 --> 00:06:23.439
Um can you define what the WoWSA streaming engine is and how it works?
00:06:24.399 --> 00:06:29.920
Yeah, I usually I usually tell people what you know, Wowza provides streaming infrastructure for people.
00:06:30.079 --> 00:06:36.480
So WoWza streaming engine is this very sophisticated kind of end-to-end platform.
00:06:36.639 --> 00:06:45.680
So we can take video in from a variety of sources, whether it be a camera, a file, um, a satellite link, whatever it might be.
00:06:45.920 --> 00:06:54.720
And we can take that video and we can we can process it, we can package it for delivery to any device anywhere in near real time.
00:06:55.120 --> 00:06:58.319
And then we can also do analysis on it along the way.
00:06:58.399 --> 00:07:06.879
Um, we can make sure that video is secure, we can encrypt it if it needs to be encrypted, we can make sure only the people who are supposed to be watching the video are supposed to be watching the video.
00:07:07.199 --> 00:07:17.519
And it's anywhere from, you know, hundreds of inputs to very few people watching, or machines watching, or single inputs to millions of people watching, right?
00:07:17.680 --> 00:07:18.959
It scales both ways.
00:07:19.040 --> 00:07:24.560
And Streaming Engine was built to be this kind of general purpose workhorse in the middle of all that.
00:07:24.800 --> 00:07:34.079
Um it's very API driven, so it's often built into other products, like the folks at IVS have used with Vault.
00:07:34.240 --> 00:07:35.759
Is the engine is built in.
00:07:35.920 --> 00:07:38.399
You know, it's part of the, it's part of the product.
00:07:38.560 --> 00:07:41.439
It's not necessarily used by itself.
00:07:41.519 --> 00:07:47.120
It's it's part of the offering and it's built into the to the product at a very deep level.
00:07:48.079 --> 00:07:53.759
And one of the questions a lot of our clients ask is how scalable is Vault?
00:07:54.639 --> 00:08:01.120
And that scalability really is built on, in large part, the WoWSA streaming engine.
00:08:01.279 --> 00:08:07.439
Can you talk a little bit about the topic of scalability and how you make that possible?
00:08:08.240 --> 00:08:08.800
Sure.
00:08:09.040 --> 00:08:11.360
Um, a couple different ways, right?
00:08:11.519 --> 00:08:19.759
So streaming engine itself is built to take maximum advantage of any underlying hardware that it can.
00:08:20.319 --> 00:08:29.199
So things like video processing lend well to being processed with GPUs or VPUs to really increase the density of the video.
00:08:29.439 --> 00:08:50.960
So if you have a thousand cameras you need to process, um, depending on how you want to deploy your architecture, you can take advantage of those to make highly efficient um workflows and highly efficient hardware setups without having to have tens or hundreds of servers necessarily running.
00:08:51.200 --> 00:08:55.440
Um, I often tell people it's also a deployment flexibility, right?
00:08:55.519 --> 00:09:03.919
So streaming engine can run on-prem, it can run on little edge devices, it can run in the cloud or a combination of all of them.
00:09:04.159 --> 00:09:09.039
So I often tell people it's like, you know, what's your most important aspect?
00:09:09.279 --> 00:09:11.600
Are you looking to scale up and down a lot?
00:09:11.759 --> 00:09:18.639
You know, so we have customers who maybe have a hundred streams going all the time, but maybe on the weekends they might go up to a thousand streams.
00:09:18.960 --> 00:09:30.799
So I would probably architect their solution a little differently so it can scale up and down very quickly and efficiently, versus someone that maybe they have 10 streams that are going out to lots and lots of viewers.
00:09:31.039 --> 00:09:33.759
Maybe they need some very fixed infrastructure.
00:09:34.000 --> 00:09:47.759
So it's really dynamic and it's really dependent on customer use cases, which is, you know, honestly kind of one of the fun parts about my job is I get to see all kinds of different uses of this and often used in ways you wouldn't expect.
00:09:48.559 --> 00:09:51.120
What are some you mentioned one use case?
00:09:51.360 --> 00:09:55.919
What are some of the more interesting use cases that you've come across?
00:09:56.639 --> 00:09:57.759
Oh, tons.
00:09:58.000 --> 00:10:07.759
So um we stream live video from Formula One cars to the pits, um, both with telemetry and track data and things like that.
00:10:08.000 --> 00:10:11.679
Um, we have streamed live video from the moon.
00:10:12.080 --> 00:10:18.879
So being part of the NASA project to stream live video from Artemis was an amazing experience.
00:10:19.279 --> 00:10:22.480
So, you know, being part of things like that are great.
00:10:22.639 --> 00:10:40.480
And then, you know, products like products like yours, products like um, you know, we have customers who are doing e911 services so that operators can, you know, get real-time video information um from people who are calling 911.
00:10:40.799 --> 00:10:44.000
So it really does range all over the place.
00:10:44.480 --> 00:10:47.759
And you know, there's always something very interesting.
00:10:48.559 --> 00:10:54.320
You've mentioned some savings that was um related to the oil rig that's monitoring underwater.
00:10:54.559 --> 00:11:06.240
Um you've also talked about like the scalability of the solution and the stability that's required of a solution that's going to be supporting moon live capture, um, formula one to pit capture.
00:11:06.399 --> 00:11:11.279
Um, what are some other types of problem solving that your company is doing?
00:11:11.440 --> 00:11:17.519
We see a lot of overlap with those sentences I just uttered, the reliability, scalability, cost effectiveness in our solution.
00:11:17.679 --> 00:11:28.960
Um, just curious if there's anything else from a from a uh presentation perspective that's helpful for prospects to know that um your team does exceptionally well.
00:11:30.720 --> 00:11:40.639
Well, so with the new tooling and all the the AI capabilities, what we're really talking about is trying learning to treat video as a data source, right?
00:11:40.720 --> 00:11:48.799
So, how can we treat video as data and really make that data in that video as effective as possible for actions?
00:11:49.679 --> 00:11:58.240
So making video an actionable resource where perhaps it was more of a passive or non-real time before.
00:11:58.480 --> 00:12:07.840
Um, I would say our team has been really, you know, on the engineering and product side, we've been really leaning into that to look at really interesting use cases for that.
00:12:08.000 --> 00:12:22.080
On our implementation and support side, it's it's really that feedback from the customers, how can these little things we're doing add up into big changes for a customer's outcome?
00:12:22.559 --> 00:12:43.600
And I think you know, the overarching scheme of that is really how can we how can we turn this into a first class data source where um perhaps in the past it was you know it was just missing that component.
00:12:44.399 --> 00:12:45.360
It's interesting.
00:12:45.600 --> 00:12:58.799
We um, you know, we spend a lot of time in the healthcare market working with like nursing schools, medical schools, um, where they're doing simulation.
00:12:58.960 --> 00:13:19.600
And one of the big drivers in that marketplace is collecting video as data to support an evidence-based framework, so that evidence being able to really um show learner competency in a particular field of discipline.
00:13:20.000 --> 00:13:32.320
So I'm curious from from an analytics perspective, what how WoWza may be looking at that kind of application as well.
00:13:32.799 --> 00:13:38.559
I think as long as you can put heuristics around what you're looking for, right?
00:13:38.799 --> 00:13:45.679
So, what determines that competency and what does that look like in a video scenario?
00:13:46.320 --> 00:13:58.320
Um, I think then you could probably, you know, we don't pretend to have every model needed for everybody's video.
00:13:58.799 --> 00:14:08.799
We designed an open platform for people to be able to add models or train models or tune models to really meet their precise needs.
00:14:09.039 --> 00:14:16.639
So I can envision somebody coming up with um something that's tuned to those clinical settings.
00:14:16.879 --> 00:14:25.360
And maybe it's okay, if you notice that A happened within 10 seconds, you should be expecting B to happen.
00:14:25.759 --> 00:14:33.679
And if you don't see B, maybe you do some sort of alert or some sort of a notification, or even something as simple as marking the time in the video.
00:14:34.240 --> 00:14:46.799
And then you start to build that over time, and you can take that data back in your models and keep training them for even, you know, more input and more correctness, if you will.
00:14:46.960 --> 00:14:57.279
Um, so really it's gonna be about the ability to again see into the video and and notice things like that.
00:14:58.159 --> 00:15:01.759
So I have a question when it's related to healthcare.
00:15:01.840 --> 00:15:04.720
Just you mentioned a healthcare use case, as Mike also alluded to.
00:15:04.960 --> 00:15:06.639
Healthcare is an industry we both serve.
00:15:06.799 --> 00:15:08.720
We have law enforcement research.
00:15:09.200 --> 00:15:11.039
These are really sensitive spaces.
00:15:11.279 --> 00:15:17.039
Um, if they're interested in using video AI, what is different about how we serve as partners?
00:15:17.120 --> 00:15:22.720
Um, or how WoW's a service as partners when they're looking to use video AI in those sensitive industries?
00:15:23.120 --> 00:15:29.440
Yeah, I think there's a couple things, is we're not sending the video off to the cloud to be analyzed.
00:15:29.759 --> 00:15:36.799
So with Streaming Engine and the VIP product, which is the video intelligence framework, it all runs on-prem.
00:15:36.960 --> 00:15:40.559
It can run on the same devices you're running your videos now.
00:15:40.720 --> 00:15:43.840
You know, you don't have to go in and rip and replace a bunch of stuff.
00:15:44.000 --> 00:15:50.799
You're not sending video off to a cloud service where you lose control of it, where it gets parked in a cloud environment somewhere.
00:15:51.039 --> 00:15:56.080
This can all happen, you know, in real time, on the edge or on-prem.
00:15:56.399 --> 00:16:00.240
And you know, Streaming Engine was meant to be deployed in those kind of environments, right?
00:16:00.320 --> 00:16:05.279
It's it gets built in, it can be deployed in a hospital within the hospital's network.
00:16:05.360 --> 00:16:26.559
It can deploy in, you know, we actually are deployed in police stations and places like that where, you know, not to mention government and other secure installations, where you know, those are private networks, you architect solutions such that you have the resources necessary to handle the scale they need to handle at.
00:16:26.639 --> 00:16:38.480
And that could be anywhere from a single small edge machine or a laptop on up to multiple servers and then you know, multiple GPUs to handle both the AI and the video processing.
00:16:38.720 --> 00:16:47.759
So it's it's really nice that you don't have to leverage a cloud service or a SaaS platform to get this.
00:16:48.480 --> 00:16:57.360
I know AI is everybody that we're talking to, that's usually their number one question is how are you using AI?
00:16:57.519 --> 00:17:02.159
And I'm kind of curious because it it's evolving so rapidly.
00:17:03.360 --> 00:17:03.840
For sure.
00:17:04.400 --> 00:17:12.640
What's Wows's roadmap like a year to even five years down the road in implementing AI in your technology?
00:17:13.039 --> 00:17:19.119
I think there's a few things, and you're you're starting to see an inkling of what some of those can be with the product re-release.
00:17:19.200 --> 00:17:23.200
So, you know, we started with relatively straightforward object detection.
00:17:23.519 --> 00:17:26.880
So, you know, hey, look at a video, there's cars there.
00:17:27.039 --> 00:17:29.920
Great, you can count the cars, whatever it is.
00:17:30.079 --> 00:17:37.599
Um, then you move to things like, okay, now I see cars going both directions on a highway.
00:17:37.759 --> 00:17:39.039
Can I count them?
00:17:39.359 --> 00:17:42.880
Now can I tell you if the cars are going the right direction or not?
00:17:43.200 --> 00:17:48.880
And then you're starting to get into even more dynamic things like, okay, now I see a bunch of cars on a highway.
00:17:49.839 --> 00:17:58.000
Can I predict if something's going to happen, whether it be a traffic jam or an accident or something like that, based on the scenes of data I'm seeing?
00:17:58.559 --> 00:18:09.119
And then you can you can move into the reason we designed the framework like we did is because we want to be able to easily add these new models and these new capabilities as they evolve.
00:18:09.359 --> 00:18:15.839
So one of the models we did work with was a model from NVIDIA that is all about synthetic video detection.
00:18:16.480 --> 00:18:21.440
So what you're starting to see is obviously more and more and more AI generated video.
00:18:22.160 --> 00:18:31.039
And from a perspective of you know news, social media, whatever it all, you might want to be able to detect that.
00:18:31.279 --> 00:18:36.400
That doesn't necessarily mean that synthetic generated video is bad.
00:18:36.640 --> 00:18:41.680
It's just you want to be able to know if it's synthetic or if it's real and take the appropriate action.
00:18:41.839 --> 00:18:43.599
You know, a social media site may not care.
00:18:43.680 --> 00:18:49.680
It's like they may just label a day, I hey, this is you know, an AI generated video reading something.
00:18:50.000 --> 00:18:55.039
Um, but a news, you know, a news, a news source might not want that.
00:18:55.200 --> 00:19:01.200
They might want to make sure that they see if it's synthetic video, they do something completely different.
00:19:01.359 --> 00:19:03.519
So things like that will keep evolving.
00:19:03.759 --> 00:19:13.680
And then you'll be able, as the models get more powerful, and you'll be able to do more things across scenes of video, right?
00:19:13.759 --> 00:19:19.759
You can already do things like context detection, and that's good for a bunch of different reasons.
00:19:19.839 --> 00:19:31.519
But what it's what it's starting to be more commonly used for is making sure that an upcoming advertisement in a video stream isn't wildly different than the content before it.
00:19:31.599 --> 00:19:38.960
You know, you don't want a happy ad with a sad, you know, you see sad scene in the movie, all of a sudden you see puppies.
00:19:39.200 --> 00:19:43.759
Um you do, but but you know, you you at least want to know.
00:19:44.000 --> 00:19:46.319
Um so things like that.
00:19:46.960 --> 00:19:49.839
You're starting to see this go a lot of different directions.
00:19:49.920 --> 00:20:01.039
But I think some of the early wins are certainly the kind of chaining of events together to determine outcomes or if an outcome may happen and some predictions.
00:20:01.519 --> 00:20:10.480
And also, you know, the there's huge problems with content moderation across user-generated content services and social media.
00:20:10.799 --> 00:20:13.920
So better uh better applying AI to that.
00:20:14.160 --> 00:20:18.799
Um so you don't have to just solve the problem with you know teams of people.
00:20:19.599 --> 00:20:24.240
Uh, because to be honest, you know, nobody really wants to watch that video in the first place.
00:20:24.400 --> 00:20:33.039
And certainly by just teams of teams of people, you're not only subjecting them to having to watch that video, but they they miss stuff.
00:20:33.440 --> 00:20:36.640
I like everything you just mentioned from a from an ethical perspective.
00:20:36.720 --> 00:20:46.480
It sounds very um where wherever you sit on the spectrum of your AI approval, everything that um you just mentioned is all like with good intent and to do good.
00:20:46.640 --> 00:20:49.119
So we hope, right?
00:20:49.359 --> 00:20:49.839
Yeah.
00:20:50.960 --> 00:20:54.960
Hey Barry, what makes the Wowza streaming engine unique?
00:20:56.319 --> 00:21:02.880
I think, well, so Streaming Engine is is kind of the OG of streaming engines, right?
00:21:02.960 --> 00:21:05.440
We launched our first product in 2007.
00:21:06.000 --> 00:21:14.960
And you know, from the ground up, it was built to be customized and added to by users, right?
00:21:15.039 --> 00:21:22.799
So it's got a full API for you to not only for command and control, but also for extensibility.
00:21:23.200 --> 00:21:31.599
So earlier this year, as an example, um, I decided I wanted to learn how to use Cloud to write code.
00:21:31.680 --> 00:21:35.200
Yeah, I wrote code all my life, but I haven't written a lot of code in the last 10 years.
00:21:35.359 --> 00:21:38.079
So I was like, well, I'll see what Cloud can help me do.
00:21:38.319 --> 00:21:52.079
And I was able to add um a new streaming protocol to Streaming Engine over the course of really just a couple of weeks using engine APIs only.
00:21:52.160 --> 00:21:59.759
I didn't have to touch any engine code, but I was able to add the protocols called mock, and it's kind of an up and coming streaming protocol for.
00:22:00.079 --> 00:22:06.720
Low latency, potentially to replace WebRTC, but uh potentially replace a bunch of a bunch of other protocols.
00:22:06.960 --> 00:22:33.839
Um but the fact that I was able to do that without having to change any code in engine itself gives you a glimpse into the other how powerful those APIs are that you could truly integrate engine very deeply into whatever workflow you need to integrate it into without having to you know worry about backward compatibility or or anything changing.
00:22:33.920 --> 00:22:36.880
And you know, you can control that code moving forward.
00:22:37.680 --> 00:22:42.000
You mentioned Web RTC, I'm which kind of piqued my curiosity.
00:22:42.480 --> 00:22:46.960
How is mock different than Web RTC?
00:22:47.920 --> 00:22:49.680
We might not have enough time for this.
00:22:49.839 --> 00:22:53.200
Uh we can use claude to summarize.
00:22:54.319 --> 00:23:14.400
Mock is a a new streaming protocol kind of designed from the ground up to take to take advantage of some newer protocols like what's called QIC, which is part of HTTP3, which is basically a UDP-based protocol that can be used in the browser.
00:23:14.799 --> 00:23:17.839
And there's significant advantages to that for streaming.
00:23:18.079 --> 00:23:25.119
TCP has a lot of overhead, um, and it builds latency when there's when there's packet loss because it does retransmissions.
00:23:25.680 --> 00:23:35.359
And you know, traditional streaming right now, we have you know, we have HLS and Dash, which are the most common things for one-to-many streaming, right?
00:23:35.599 --> 00:23:41.599
If you're watching Netflix, if you're watching Apple TV, whatever it is, it's going to be HLS or Dash.
00:23:41.920 --> 00:23:47.200
Um, WebRTC was originally designed for video conferencing, still does a pretty good job of that.
00:23:47.359 --> 00:23:49.200
It's not as great at one-to-many.
00:23:49.359 --> 00:23:51.839
Um, it has problems in other areas.
00:23:52.079 --> 00:23:57.519
It wasn't really designed for things like accessibility, and it wasn't designed for things like DRM.
00:23:58.000 --> 00:24:10.960
So basically what Media Overquick is is attempting to do is take a new look at streaming based on the technology we know now, rather than evolved over the last 10 or 15 years.
00:24:11.200 --> 00:24:13.759
It's like, how would we do this if we were starting over?
00:24:14.480 --> 00:24:16.960
And kind of trying to be the best of all worlds.
00:24:17.039 --> 00:24:18.319
And it's got a lot of cool features.
00:24:18.480 --> 00:24:20.880
You can do super low latency with it.
00:24:21.039 --> 00:24:23.839
Um, you know, it's it's playable in browsers.
00:24:23.920 --> 00:24:28.559
It's you can broadcast from a browser, but you can also broadcast from a camera.
00:24:28.880 --> 00:24:31.039
Um, it scales really well.
00:24:31.200 --> 00:24:34.640
You can kind of tune the latency for your audience size.
00:24:35.039 --> 00:24:45.119
It's designed to be deployed with what we call relays, which you send a stream to a relay, that relay can then send streams to people or other relays.
00:24:45.279 --> 00:24:48.000
So you get this kind of treed fan out effect.
00:24:48.319 --> 00:24:50.240
So there's a lot of promise there.
00:24:50.319 --> 00:24:51.839
Um it's getting close.
00:24:52.160 --> 00:24:56.559
You can certainly go um read lots about it.
00:24:56.720 --> 00:25:05.920
Open mock is an uh industry group that's kind of deploying versions of it as the spec evolves.
00:25:06.160 --> 00:25:15.920
Um, but I think it's going to be a really interesting, interesting alternative to some of the other um protocols over the next couple years.
00:25:16.319 --> 00:25:33.759
What kind of guidance and recommendations can you offer organizations that may either be looking to upgrade some of their video streaming capabilities or just getting started out in terms of, yep, we recognize a need uh to stream video.
00:25:34.000 --> 00:25:41.440
It could be a nursing simulation that wants to stream from a live simulation to a classroom of students, for example.
00:25:41.599 --> 00:25:46.319
Um whether that classroom of students is across the hall or across the country.
00:25:46.640 --> 00:25:47.039
Yep.
00:25:47.359 --> 00:25:54.480
And they're kind of getting started, like, well, you know, the big question could be where and how do we get started?
00:25:55.759 --> 00:26:00.960
I my first advice would would, you know, start small, right?
00:26:01.200 --> 00:26:09.759
So pick one clear use case, one workflow, one problem that you're trying to solve, and get it working end to end, and get it working well.
00:26:10.240 --> 00:26:11.119
We'll take your case.
00:26:11.200 --> 00:26:19.119
You know, I have I have one class I'm trying to stream from one place to another.
00:26:19.759 --> 00:26:30.400
Get that working, get the recording working, get the interactivity elements working, get that dialed in, and then you will you'll understand what you need to do that at scale.
00:26:31.200 --> 00:26:41.680
So, you know, I would say don't try to jump in and do 200 classes next week to you know 200 different sites.
00:26:42.160 --> 00:26:44.000
Um get it working.
00:26:44.319 --> 00:26:50.079
And then many times you have the infrastructure already.
00:26:50.240 --> 00:26:52.400
You have cameras, you may have servers.
00:26:52.480 --> 00:26:58.480
You don't necessarily need to go out and buy a whole broadcast studio's worth of infrastructure to get started.
00:26:58.720 --> 00:27:03.920
You know, start with what you have because you know it may be more than good enough to get you there.
00:27:04.240 --> 00:27:09.599
Um and then, you know, find people to help you, right?
00:27:09.759 --> 00:27:23.279
I mean, whether it be vendors or whether it be communities, but there's a lot of there's a lot of places to get advice on you know representative best practices and how to do things.
00:27:23.440 --> 00:27:30.720
I'm not I'm not gonna say the right way to do things because that's different for every implementation and different for every person's use case.
00:27:30.799 --> 00:27:36.799
But there's a lot of yeah, there's a lot of resources out there for you to take advantage of.
00:27:37.039 --> 00:27:40.160
And that's where I think you know, Wowza certainly is great.
00:27:40.240 --> 00:27:43.359
We have a great support team, we have a great services team.
00:27:43.519 --> 00:27:48.720
They understand not only Wowza's products, but they understand a lot of other products and how things all work together.
00:27:48.799 --> 00:27:50.400
And I think that's important too.
00:27:50.720 --> 00:27:52.480
So you mentioned infrastructure.
00:27:52.880 --> 00:28:01.279
I'm kind of curious what kind of infrastructure requirements are you looking for in order to kind of deploy your solutions?
00:28:02.640 --> 00:28:04.640
So we have a ton of flexibility in that regard.
00:28:05.359 --> 00:28:12.400
As I mentioned, we can deploy on tiny little edge servers, you know, not much bigger than a Raspberry Pi, to be honest.
00:28:12.480 --> 00:28:14.960
They're very, very small little devices.
00:28:15.200 --> 00:28:21.440
And if you only have a camera or two, you're looking to get repackaged and streamed out for mobile consumption or whatever it is.
00:28:21.680 --> 00:28:22.880
That may be enough.
00:28:23.279 --> 00:28:29.440
And if you have a remote setting where you're trying to do something, maybe plunking one of those down in that remote setting is gonna be good.
00:28:29.920 --> 00:28:34.319
And you know, you have options as far as what your input source might be.
00:28:34.400 --> 00:28:46.160
You could you could be streaming from a phone into that box, you could be streaming from a you know a dedicated camera or even a professional camera, and you know, you can start there.
00:28:46.400 --> 00:28:53.920
And you know, typically people are these days is almost most streams are going to be viewed um on a mobile device.
00:28:54.640 --> 00:29:01.279
So optimize your stream so it works well on a mobile device, it works well under all different bandwidth conditions.
00:29:01.440 --> 00:29:11.200
You know, and streaming engine is great at creating sets of outputs that look good on mobile devices under varying bandwidth conditions and varying screen sizes and things like that.
00:29:11.519 --> 00:29:12.720
I like that advice you gave.
00:29:12.799 --> 00:29:17.759
And I'm curious what type of uh internal organizational context do you work with?
00:29:18.079 --> 00:29:25.119
Do you work with a lot of end users or do you work with the network team or who are you really collaborating with from the ground up?
00:29:25.519 --> 00:29:27.839
So it's it's kind of a combination, right?
00:29:28.079 --> 00:29:38.480
So I will typically talk to the business owners of the problem to understand what it is they're trying to accomplish.
00:29:38.799 --> 00:30:02.559
But then when it comes to architecture and implementation, it's it's someone in the technical team, whether whether they have a video engineering group or whether it's the IT team or a combination of the two, um, will work with them to understand their infrastructure constraints or opportunities on the deployment side and how to set up a solution.
00:30:02.640 --> 00:30:07.440
And that's where we often also will say, you know, bring in other vendors.
00:30:07.599 --> 00:30:11.759
You know, you may need a delivery provider to deliver things at scale.
00:30:12.079 --> 00:30:18.160
You may need, you may need a cloud provider at some point to step in to do massive storage or something.
00:30:18.400 --> 00:30:27.039
So understanding where things are and what things needs to happen will also influence um the kinds of partners you might need to bring into the solution.
00:30:27.359 --> 00:30:32.240
And you help with that overall design so you can help make rough like referrals or you can help uh guide.
00:30:34.160 --> 00:30:40.079
We we have tons of partners in the industry, not only from like a systems integration perspective, but vendors.
00:30:40.160 --> 00:30:46.319
Like if we if we need a DRM vendor, we can bring in a DRM vendor we've worked with in the past that we know is going to work.
00:30:46.559 --> 00:30:50.319
Um, encoders, cameras, you know, CDNs.
00:30:50.880 --> 00:31:01.680
Um so we definitely have the whole aspects of the ecosystem covered as far as how we attack problems.
00:31:03.839 --> 00:31:12.880
So, Bear, before we wrap up, is there anything else that you'd like to share with our audience or questions that we should have asked that we missed?
00:31:15.599 --> 00:31:16.480
Let's see.
00:31:16.799 --> 00:31:25.440
Um I think let me look through my notes.
00:31:25.680 --> 00:31:31.519
Um I think, I mean, we we covered a wide range of ground, and I think it's interesting.
00:31:31.599 --> 00:31:54.000
And I think, you know, what I would what I would you know say if you're streaming now, um take a look how you're actually using that video, and if some of the new things coming out in terms of of AI and and video intelligence might help you.
00:31:54.240 --> 00:31:57.920
Um the answer may be not right now, and that's fine.
00:31:58.160 --> 00:32:20.240
But you know, I think what you'll find is that there's data buried in that video, and you're really looking to make the most out of your infrastructure and your video investments, building that case for, hey, we've done great with video now, but look at all this stuff that's actually opening up and we can we can look at and do some experiments, right?
00:32:20.319 --> 00:32:24.079
I mean, it's um whatever it might be.
00:32:25.599 --> 00:32:27.519
Well, that's awesome advice.
00:32:28.079 --> 00:32:34.480
How can our listeners get in touch with you if they they want to have a deeper conversation?
00:32:35.839 --> 00:32:37.759
Well, they can certainly find me on LinkedIn.
00:32:38.000 --> 00:32:39.359
Pretty easy to find there.
00:32:39.519 --> 00:32:40.880
It's Barry J.
00:32:41.039 --> 00:32:43.519
Owen, whatever the whole LinkedIn thing Barry J.
00:32:43.680 --> 00:32:44.559
Owen is.
00:32:44.960 --> 00:32:47.759
Um, you can find me on waza.com.
00:32:47.839 --> 00:32:49.839
I write a lot of blog posts there.
00:32:50.480 --> 00:32:59.359
And you know, occasionally you will see me speaking at events or at conferences, certainly at trade shows.
00:32:59.519 --> 00:33:01.839
Um, but I'd love to hear from people.
00:33:02.079 --> 00:33:04.799
Happy to connect on LinkedIn and talk more.
00:33:05.279 --> 00:33:09.519
And you know, I'm not going to come at you with a hard sell.
00:33:10.000 --> 00:33:13.279
I, you know, you must buy WoWza because of this.
00:33:13.440 --> 00:33:22.160
Um, I really want to come at you more of what problem are you trying to solve and what's the right way to go about it and how you know how are you going to get the most out of what you're doing?
00:33:22.480 --> 00:33:23.680
What a great resource.
00:33:23.920 --> 00:33:24.240
Yeah.
00:33:24.480 --> 00:33:25.759
Well, thanks for your time, Barry.
00:33:25.839 --> 00:33:26.880
We really appreciate it.
00:33:26.960 --> 00:33:32.640
Um, I think our listeners have uh will learn a lot uh from our our podcast.
00:33:33.200 --> 00:33:33.920
Well, I hope so.
00:33:34.079 --> 00:33:34.559
It's fun.
00:33:34.720 --> 00:33:35.680
Appreciate the time.
00:33:35.839 --> 00:33:36.720
Thank you.
00:33:38.559 --> 00:33:46.160
OnRecord is a podcast presented by Intelligent Video Solutions, hosted by Mike Anzolone and Carmen Hawk, and produced by Kyle Shellstead.
00:33:46.319 --> 00:33:53.440
Find us online at ipivs.com slash on record and on social media at Intelligent Video Solutions.
