Voices of Video
Explore the inner workings of video technology with Voices of Video: Inside the Tech. This podcast gathers industry experts and innovators to examine every facet of video technology, from decoding and encoding processes to the latest advancements in hardware versus software processing and codecs. Alongside these technical insights, we dive into practical techniques, emerging trends, and industry-shaping facts that define the future of video.
Ideal for engineers, developers, and tech enthusiasts, each episode offers hands-on advice and the in-depth knowledge you need to excel in today’s fast-evolving video landscape. Join us to master the tools, technologies, and trends driving the future of digital video.
Voices of Video
600 Channels, 1 Server - Reimagining Video Processing for Sustainability
Hardware acceleration is changing the economics of video delivery, and Scalstrm is at the forefront of this revolution. After showcasing their just-in-time transcoding solution at NAB, Dominic returns to Voices of Video to share how they've partnered with NETINT to create a groundbreaking platform that's already winning customer deployments.
• Scalstrm built their transcoding solution from scratch without relying on open-source components
• Their migration from software to hardware acceleration took only 3-4 months
• Performance testing showed "huge" advantages using VPU cards over CPU-based transcoding
• The solution supports all major codecs including H.264, HEVC, and has plans for AV1
• They demonstrated 600 channels of origin packaging plus 64 live transcoding channels on a single small server
• Their target customers include tier-one telcos, broadcasters, and content owners
• The platform now includes just-in-time transcoding, subtitle conversion, CDN capabilities, and ad insertion
• Their first live customer deployment is happening immediately following NAB
The hardware acceleration revolution is here - join us next time for more conversations with companies building real technologies and products that are changing video encoding and streaming.
Stay tuned for more in-depth insights on video technology, trends, and practical applications. Subscribe to Voices of Video: Inside the Tech for exclusive, hands-on knowledge from the experts. For more resources, visit Voices of Video.
Voices of Video.
Mark Donnigan:Voices of Video. Voices of Video. Voices of Video. Well, welcome back to another very exciting episode on Voices of Video. Now we had a whole buildup before NAB and of course, if you're watching this, you know two or three months from now, then references to NAB may not make sense. But if you were able to join us that is, netent at NAB, you perhaps saw the flurry of activity in our booth. We were joined by five partners. We did something very different this year and we brought five partners into the booth with us who were all showing their various solutions and their platforms that are integrating or where they've integrated with VPUs. So it was wonderful, it was an amazing time. We had a ton of fun and the booth was just jammed. It was jam-packed. Today I am talking to Dominic from ScaleStream, and ScaleStream joined us at NAB. So first of all, dominic, welcome back to Voices of Video.
Dominique Vosters:I'm glad to be back and also many thanks for the opportunity to join NAB at your booth. It was a really exciting time and we can continue to talk about that.
Mark Donnigan:Yeah, yeah, it really was. You know, the whole concept this year was VPU ecosystem, and the concept behind that, which your company has played a key role in, is that when you need to integrate hardware into the network, there's obviously customers and the platforms themselves have to go through some sort of evaluation process where they just literally doesn't meet the quality bar. They have to verify that. You know run. You know various test scenarios. They have to go through some. You know just quality, what happens if this thing runs 24-7 and is never turned off? You know, does it need to be? You know, rebooted every three days, or you know. So there's all of these steps that you have to go through.
Mark Donnigan:And so we, you know, as the market for VPUs has developed over the last five years, we are very fortunate, and you know, and we actually feel quite humbled to have started with some very large hyperscalers and of course, they own their own platform. They have large engineering teams that can do the integration. So they brought our product, bought our product, put it through, you know, intense scrutiny, installed it in the network, and you know, and they use it and we have like more than a hundred thousand out there in the market today that are actually running in production 24, seven. But, as we know, you know, if you think of the tip of the pyramid, top of the pyramid, these hyperscalers, you can almost count on two or three hands. There's not so many of them. Right, they are huge, but there's not so many, and that's why we're so happy to have our partnership together with ScaleStream, that we can take this great technology, the VPU technology, and make it available to everybody, and so that was exactly the point at NAB.
Mark Donnigan:So you know, I want to start here. Dominic, you and I had a conversation. Our booth was jammed, very, very busy. I know you guys, you know I felt like every time I turned around. You know, you had a meeting and you were very busy. All of our partners were really but there was talk at NAB about traffic being down, attendance being down. Did you notice that, or were you hearing the same things, or what was your observation?
Dominique Vosters:Yeah, I think it's a good point. I think in general we noticed that there were less people I think less attendees at NAB, but then again I don't think it was a real disadvantage because I think the quality of the meetings were much higher than just people passing by. People were there with a reason, I think. So in my opinion, everybody at NAB was there with a reason and therefore I think the meetings were good. So for us we noticed that there was less people walking around, but then again at our boot we had a lot of good meetings. Some of them were pre-booked, but also people were passing by and then, yeah, they came with a reason to our boot. Either they heard it from partners or from other people, like, please jump by to the NetTint booth at 10 and skill stream and have a conversation or a demo. So for us it was not really something that was a bad thing, but indeed I think there were less people in general.
Mark Donnigan:Yeah, yeah, you know I have done. I don't even know. I was sitting there trying to think how many NABs have I done? You know it must be at least 15, which 15, which I know there's some people say, oh, come on, mark, I've done 30. But I guess I'm not so old in the industry, but you know. So certainly traffic was down. But you know, let's face it, at the end of the day we're there to do business. We are there to, you know, have, I would say, real meaningful conversations and 100 encounters where we're just giving out T-shirts. The value is not as high as those 20 or those 30, or those 40 really focused meetings. I fully agree, yes, yeah, yeah, correct, I fully agree, yes. So you had a very successful show and you know why don't you give a preview of some of the results? You know that came out of NAB for you.
Dominique Vosters:Yeah, definitely. So the main reason we were at NAB was to showcase a bit our new products, mainly on the transcoding side, and one of this was, of course, that we used the VPU card for the transcoding part. And at IBC. Actually, everything started at IBC last year, where we initially demonstrated our just-in-time transcoding solution. There was a lot of interest for that and then we started building a product around it.
Dominique Vosters:We started the discussion, I think, right after IBC, with NetInt as well.
Dominique Vosters:We built a product around it and then at NAB, I think we had a ready product that we could fully demonstrate, not only for VOD just-in-time transcoding, vod offline transcoding, but also live just-in-time transcoding.
Dominique Vosters:And then I think this was something well, let's say, relatively new that not all our customers partners knew or were aware of, because our website is not fully up to date yet. So we had really good conversations about this. People, customers, partners were really interested in this. For us this also opens up new markets that we can, for example, handle new use cases like hospitality, pop-up channels, where we can do more or less an end-to-end chain from live transcoding till the origin, till the player, let's say. And then even I must say that this was really successful because even I think tomorrow, tomorrow afternoon we already have a live event planned with a just-in-time transcoder running on VPU cards, so that's really positive. We have some other opportunities that we're discussing pricing and requirements in Asia and then another one in Europe. So it was a really successful NAB and this definitely opens doors for all the use cases and opportunities.
Mark Donnigan:So yeah, it was good, amazing. Well, congratulations. As I said there in my first question, at the end of the day, we're there.
Mark Donnigan:Yes, it's great to connect with our friends and all, but we're all there to do business, exactly so the fact that you walked away with very concrete opportunities. I want to talk about your just-in-time transcoder, because there's a couple things that I would like you to tell the audience about I'm sure they would be very curious, that I would like you to tell the audience about, I'm sure they would be very curious. One is how you went from IBC, which is September, so I guess where we sit today, call it six, eight months, but in reality you had to have everything built well before April, the 6th, the start of NAB. So presumably you were able to make this migration from software to hardware in what? Maybe three or four months. Is that how long it took?
Dominique Vosters:you yeah, more or less Correct.
Mark Donnigan:Okay, yeah. So maybe you can, you know, walk us through the journey Like what did that look like? What is some of your underlying architecture? Are you leveraging open source frameworks like FFmpeg? Have you built your own solution? I know the audience would be super curious to hear how you did that.
Dominique Vosters:So I think in general at Skillstream we built everything from scratch, so we're not using any open source components, and we try to do this in every product, more or less. Of course, sometimes you need to rely on existing libraries, but for the rest I think everything is more or less built in-house. And we do this for a reason it gives us a lot of flexibility, it gives us the best performance and then with this we can really be really agile and adopt really fast to new features, new requirements, and not rely on open source components. So we do the same for the transcoding part. So at IBC, initially everything was based on CPU. This works fine, but of course yeah, you need to have a lot of CPU power.
Mark Donnigan:You do, especially for live.
Dominique Vosters:Exactly, especially for live and for high bit rates, depending on the callc and so on. So after we started the discussion with Netint, we started testing this in-house and then, yeah, I think one of the things that is really high on our agenda is sustainability and cost efficiency and, yeah, I think the VPU cards really fit well into that strategy. Is really high on our agenda is sustainability and cost efficiency and, yeah, I think the VPU cards really fit well into that strategy because on the origin, we try to be the most performant platform of the market. Yeah, I think now with transcoding, with using the VPU cards, I think we did some benchmark tests and the performance difference was huge. So it's a no-brainer to do the transcoding in hardware on the VPU cards.
Dominique Vosters:That was quite clear from the tests and afterwards yeah, we really built the product around it.
Mark Donnigan:So, yeah, that's really good, that's great.
Dominique Vosters:What codecs are your customers primarily focused on? Yeah, it's still the classical ones H.264 and H.265. Yeah, HVC, so we get in RFPs. Av1 is typically one that is requested, Not that it's a mandatory codec but it's definitely something that's considered for the future, but for now it's still H.264, H.265. And I think that's what's commonly used.
Mark Donnigan:Yeah, no doubt. We see that really regardless almost any region of the world now that really regardless, almost any region of the world. Now it's not just in America or in certain European countries, it's everywhere. It's H.264, hevc or mandatory. And AV1 is either a nice-to-have or a requirement, though they're not using it today, you know. But they're not going to, you know, they just simply cannot ignore AV1. Av1 is coming, it is here, it's happening, you know. And so if you're going to be investing especially in hardware, you better make sure it supports AV1. And of course we have a very good solution there. So you know it always blows people away. Maybe you even had the same experience.
Mark Donnigan:We're so used to software right, and H.264 is, you know. Obviously it depends on at what time period you made the measurements in terms of the complexity of the compute required, but you know it's anywhere from four to eight times. You know. Maybe in the early days it was 10 times more complex, meaning it required more cycles, that many more cycles. Going from H.264 to HEVC, then going from HEVC to AV1 was like another almost 10x. I mean again, you know the practitioners will say no, mark, you know. Now you know the software, you know codec, engineers and all have greatly improved performance. And that absolutely is true.
Mark Donnigan:Svt-av1, you know, and there's certainly some proprietary AV1 encoders that are. But the point is it required a lot of compute. So we're used to thinking software where you could say, oh, h.264 performance is actually reasonable enough, you know. And then you go to ATVC and you say, oh, I'm going to cut my stream count by four, you know, on a server maybe even five. Then you go to AV1 and you're like I'm lucky if I can do one or two ABR ladders on even a massive, you know, big AMD, amd 4864 core processor. But with VPUs it's the same. So you get 32, 1080p, 30s, whether on a single chip solution like a T1U, which I believe is what you guys are using, you get 32, whether it's H.264, hevc or AV1. People are like wow that's just amazing.
Mark Donnigan:They can't believe it.
Dominique Vosters:And this was also really mind-blowing because at NAB we had a kind of demo server with us, because we can't really rely on the internet to do demonstrations.
Dominique Vosters:So we had our own demo server with us, because we can't really rely on the internet to do demonstrations. We had our own demo server with the VPU cards in it and then people were like, yeah, how many channels can you do? And we just opened the door from the closet and we showed our demo server, which was a really small form factor, and on this one we could do 600 channels origin repackaging, and then I think we had two cards so we could do 64 live channels transcoding just on that device. And this was really mind-blowing for a lot of people, like, oh wow.
Mark Donnigan:Yeah, oh, wow. Yeah, that's incredible. Yeah, that's really great. Well, where do you see you are primarily working in I'll call it broadcast applications? Is that correct? Or where is your sweet spot in terms of what your focus is, across the industry or for the industry?
Dominique Vosters:Yeah, Our main customers are a few different types. Everything started with the tier one telcos, which still is, of course, a clear focus for us, but then, I think roughly two years ago or perhaps even three years ago, we also included included broadcasters in our customer's portfolio, and then now, with just-in-time transcoding, we have even more use cases. Like I mentioned, hospitality, pop-up channels, educational purposes, webinars that we can handle, but I think the main customer focus at the moment, our main customers, are currently the tier one telcos and broadcasters and partly also the content owners with just VOD, but this is mainly on VOD transcoding and packaging.
Mark Donnigan:Yeah, got it, got it. And so are these telcos that are currently offering entertainment channels, or are they operating on behalf of some of their large media customers? Or you know? Maybe you can tell more about what these workflows are. You know who the end customer is, what the content is they're delivering?
Dominique Vosters:Yeah, I think it's a bit of a combination. Either tier ones, of course, they have their own customers, either with set-up boxes or smart cars or even just an app, and then often they also resell part of their channels to third parties. So this is something that they can do with our multi-tenant platform, so they can just split off a few channels and resell them to third parties, and then they also act as a kind of aggregator, because you have all the individual broadcasters within different countries and they kind of aggregate all these local public commercial channels and they offer it to all the customers as one package.
Mark Donnigan:That's right, yeah, yeah, interesting. Well, it's a very important business. It's a good business. You've designed a very good solution and, like I say, I was impressed. It really did feel like every time I had a little break and kind of was looking around, you were hard at it in the middle of a conversation you know so with customers. So there's a lot of interest clearly in the platform. Yeah, definitely. So where do you go from here? Maybe you can give us a little sneak peek. You don't have to give away all the secrets, but how is the platform going to be developed? Is there anything new or anything that you're going to add for IBC? Maybe you can give us a little preview.
Dominique Vosters:Yeah, I think. First of all, yeah, at NAB we demonstrated our solution. We started at IBC the-in-time transcoding, so now we have a ready product. We are having the first discussions with customers. Tomorrow we even have a live event. It's a proof of concept, but actually it's a live proof of concept because it's with real customers and a real live channel. So I guess it's really good.
Mark Donnigan:Yeah, that's always good when it's live and it's real. You know that's.
Dominique Vosters:Exactly.
Dominique Vosters:So, that means that, yeah, it's a good test for us. Of course, yeah, we will have continuous discussions with other customers to see where we which requirements we need to fulfill. And this is, I think, also where I vote ScaleStream and NetInt to work closely together our engineering team, your engineering team that we get feedback from customers and partners like, yeah, we need this feature or that feature that we can implement quite fast. So we add more and more features to it. It also fits well into our server-side ad insertion, server-guided ad insertion solution, because also there the ads need to be transcoded. So I think it fits well into our product portfolio and it makes our solution even more end-to-end, I would say.
Mark Donnigan:Yeah, for sure. So you know it's interesting. We didn't talk about your ad solution. Maybe, as we wrap up here, why don't you go through and give an overview of the entire platform, what's in it, you know? Say a little blurb about each, you know, each functional block?
Dominique Vosters:Yeah, perfect. So I think everything started with the origin just-in-time repackaging and this was our first product. Let's say three focus points. First of all, performance this was our first product. Let's say three focus points. First of all, performance. This was something from the start so we could do 600 channels on one rack unit server. I think we discussed it last time. We replaced 40 or 48 servers with just two. So that's a real key differentiator Ease of use and flexibility that we can run in the cloud and that is easy to operate.
Dominique Vosters:Just the I2T image-to-text subtitle conversion. This is kind of an add-on where we can convert TVB image subtitles to text. Later on we added our Silt and CDN on-prem CDN and Silt, which was added as a product, and then later on we added offline transcoding and then recently we added just-in-time transcoding with VOD and live as well at the moment. And then of course it's not really a product on its own but it's more a feature or an add-on. The ad insertion flow in general and of course server-side ad insertion is something that is a must-have, but also there we see that the market is evolving or at least exploring server-guided ad insertion, which gives a lot more flexibility and probably also more quality and then possibly more revenue for the full chain. I think so this is currently our main product portfolio, but of course we're always open to listen to our customers, partners. If we're missing some functionality, then we're happy to have the discussion and discuss if we can add new stuff as well.
Mark Donnigan:Yeah, yeah, absolutely. How much of the solution that you bring to a telco is the product you know I'm going to say out of the box, and then how much do you need to customize, like do you find there's always? You know I'm going to say out of the box, and then how much do you need to customize, like do you find there's always, you know, 10%, 20% you have to customize. Or are you basically bringing the product you know with minimal customization?
Dominique Vosters:Yeah, good question. We try, of course, to be less, not to customize per customer because it needs to be scalable.
Dominique Vosters:So software-wise, we definitely only have one software stack so we don't customize per customer. Also, our origin has proven itself in different environments tier one, telcos, broadcasters. So I think in 99% of the cases all the features are there and it's just some fine tuning that we might need to do for specific use cases. So, yeah, I would say mostly it's plug and play with some small customizations, but it's not really a customization. It's more tweaking or fine tuning depending on the use case. Yeah, yeah. Yeah.
Mark Donnigan:I think you know everybody would like to be a product company from the perspective that it's a full out of the box solution. Right Video is just. The requirements are just too diverse. It just isn't really possible. We all are delivering solutions where you know you hope to bring as close to 100% as possible to every single customer. But it just, you know, it's just not practical or or realistic.
Dominique Vosters:So yeah, no, I mean, there are always special use cases, yeah, yeah.
Mark Donnigan:But? But I always like asking the question because there's certainly nothing wrong with with a bespoke engineering type approach, and you know, there's even some customers who maybe that's exactly what they want. You know is to come, let's code, let's design this together and then, you know, let's co-create the solution. But I sometimes smile when I see companies representing themselves as a products company and it's like wait a second, you're pretty much building every solution new, so it's just a different approach, that's it.
Dominique Vosters:I think in the end it's a good thing, because in the end, all our customers will benefit from it because if you need to do some fine tuning for one customer, other customers can benefit from it.
Mark Donnigan:And that's exactly right. You know that's the benefit of a product that has years and years and years of development behind it. You know it's the same with us. We have, you know, we have a wide range of applications, of use cases of you know. I would even dare to say sophistication, you know, and that's not to say that you know, some customers are smarter than others. That's not at all what I'm saying. But you know, the fact is there. You know, we have some customers where the encoding team is 30, 40, 50 people, you know, and all they're doing is working at the deepest levels of the codec, you know. And there's other teams that you know. It's one person who you know maybe can carve off five hours a week to go do some evaluations. You know, it's just how the company's built right. So the beautiful thing is, you know the refinements and the enhancements that we make for the team of 50, you know, deep codec engineers, phds, well, guess what? All of our customers, generally speaking. You know, sometimes there's features that you know we're not able to roll out to everybody, or just when it makes sense, but everybody benefits. So it's yeah, yeah, so it's definitely.
Mark Donnigan:You know, like I think, of CRF, you know and this was one of the things we promoted we introduced CRF now, oh, a year and a half ago In fact, actually, maybe it was two years ago the ability to make a change, add a feature like CRF to the encoder, to a hardware encoder, was revolutionary. People were like you can modify the rate control? It's like, yes, and after we actually shipped the product, so every VPU in the field was able to be upgraded, both Logan and Quadra. Logan was our first generation, but with CRF, you know, rate control absolutely amazing. So you know again, something that's been available for many, many, many, many years on, like X264, x265.
Mark Donnigan:But the fact that you could add that to hardware was, you know, really, yeah, really, really, really revolutionary and opened a lot of people's eyes, you know, to the flexibility that we offer. Well, dominic, thank you for circling back around. You know this is our second Voices of Video interview, you know, leading up to NAB, and then now we get to close and kind of say, here's what we saw. And you know, here's the progress, the success. We will definitely do something similar before IBC and we're going to be talking more. So I wish you the best, Thank you for your support and we're really happy to have ScaleStream be a part of our ecosystem, the VPU ecosystem.
Dominique Vosters:Yeah, thanks a lot and we also really appreciate the opportunity to be at NAB at your booth and we really appreciate the partnership as well and looking forward to the next steps in the coming months, Absolutely.
Mark Donnigan:Yeah, great, Super, Okay. Well, thank you to all the listeners. We really do appreciate your time and the attention that you give us. Voices of Video. We strive to bring real conversations with companies and people who are building real technologies, real products. They're doing the real stuff. You're not going to get a lot of sort of industry analysis and, you know, debate and discussion about things that you know sometimes are sort of like the weather there's absolutely nothing that we can do to change, whether we like it or not. So what we strive to do here at Voices of Video is bring the real facts, educate you, hopefully entertain you a little bit, but help improve your knowledge of what's happening in video encoding, video streaming, bring along the technology advancements like VPU, and we hope this is helpful. So until the next episode, as we commonly say, happy encoding.
