Synchronizing 20 Perspectives: The Future Of Multi-View Esports

Cameras miss moments; fans don’t. We wanted every decisive peek, every clutch revive, and every chaotic final ring in Apex Legends to be watchable from any team’s perspective - live, synchronized, and affordable. That meant rethinking how we transcode and distribute dozens of POV streams at once without drowning in startup lag or compute spend.

We walk through how Scalstrm integrated NETINT VPUs at a low level to pack up to 20 live channels onto a single card, slashing both costs and boot times for event-based streaming. Instead of relying on generic wrappers, they tapped direct APIs to tune buffer behavior, rate control, and ABR ladders for fast-motion gameplay.

Partnering in the Akamai Cloud lets them spin up encoders only when needed, bring them online in seconds, and tear them down post-show—no idle fleets, no waste. For VOD, just-in-time transcoding stores a single high-bitrate master and generates renditions only when requested, keeping catalogs lean while preserving quality.

Znipe Esports takes the spotlight with a multi-POV esports product that delivers 20+ synchronized streams plus the main event feed. To keep every angle aligned, they apply AI and image analysis to lock onto in-game clocks, then validate with operators for frame-accurate sync across teams. Telemetry from damage and kill events fuels real-time overlays and instant highlights, so fans can jump to the best moments or follow their favorite squad without missing context.

The payoff is dramatic: 25% lower transcoding cost, 70% faster startup, and a 75% reduction in high-quality transcoding cost—exactly where esports audiences are most demanding.

We also share a war story: going live in 30 minutes only to find GPU capacity swallowed by AI training. VPUs gave us a dedicated path for video, restoring predictability when it mattered most.

If you care about multi-view control, synchronized angles, and high frame-rate streams that don’t blow up your budget, this breakdown shows how to get there.

Listen now: https://netint.biz/podcast
Download the presentation: https://info.netint.com/hubfs/downloads/VPUs-on-Akamai-cloud.pdf
Test drive NETINT VPUs on Akamai Cloud and get $500 credit: https://netint.biz/akamai_500

Episode highlights:
• Scalstrm’s origins in packaging, origin, and analytics for operators and broadcasters
• Why low-level VPU APIs beat generic wrappers for live density and efficiency
• Instant provisioning for event-based transcoding on cloud partners
• Just-in-time transcoding for VOD to cut storage and compute
• Znipe’s multi-POV product for Apex Legends with 20+ team feeds
• AI and image processing for frame-accurate sync on in-game clocks
• Ingesting telemetry to render stats and auto-generate highlights
• Cost wins: 25% lower normal transcoding, 70% faster startup, 75% lower high-quality costs
• Avoiding GPU shortages by shifting to VPUs for predictable capacity
• Higher resolutions and frame rates that match esports viewer expectations.

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.

Chapters

• 0:07 Who We Are At ScaleStream
• 0:28 Why VPUs For Live Transcoding
• 2:15 Instant Provisioning For Events
• 5:09 Enter Snipe And Esports Needs
• 6:56 Apex Legends And Multi-POV
• 9:25 Syncing Streams With AI And Telemetry
• 12:33 Why Fans Love Multi-View Control
• 12:57 Cost And Scale Challenges
• 14:21 VPU Gains: Cost, Startup, Quality
• 16:16 The GPU Capacity Crisis Story

Transcript

Voices of Video: 0:07

Voices of video, voices of video. The voices of video. Voices of video.

Pontus Eklöf, Scalstrm: 0:17

So my name is Ponto Seclev. I am a sales director at ScaleStream, and with me today I have Per Nyboum from Snipe. And we're here to talk a bit about how we at ScaleStream use VPUs in general, but more importantly for event based streaming. And Peris gonna talk a bit more about that later, but I will start with giving you a brief introduction to who we at ScaleStream are. Thär vi går. So ScaleStream. We are a Swedish based software company. We work primarily with operators and broadcasters. We're historically been fokusing on origin and packaging functionalities when it comes to ATT and streaming services. Så a lot of our kussmers, operators and broadcasters with multiple channels that they managed and need to continuously repackage content and get statistics och analytics on performance and so on. And we have built a system that is very robust, very scalable built on mikroservices so that it skales skills and easily with us. So our kustmers expect a ease of use and a high reliability with a lå need av servers. A very hög i efficient visag aven kapacet. A need to have uh efficient transcoding solutions fully integrated into our system. So we started looking at what would be the best way for us to add as a fully integrated feature the capability to transcode content as it was being ingested to our origin server. Then we came across the VPUs for net and we saw uh extreme high efficiency. Uh we could do up to uh 20 live channels on a single card uh in a much more efficient way than than we could with any other type of solution. And we built our integration on their low data APIs. So this is not just an integration using uh FFMP, but rather using the low-level uh APIs uh to to get the full benefit and full efficiency of utilizing the uh net in the use. Uh and uh when we um started out building this uh feature into the system, and we also realized that it opened up other opportunities and new possibilities uh in terms of uh how this could be deployed. So we are partnering with uh Ackerman uh as a uh QCT partner, uh, so you can get our technology available in the Glenode servers where you can utilize the VQs on the fly, uh and this opened up uh possibilities both in terms of uh deploying uh our servers really quickly but uh also to have them deployed instantly for event-based transpoding. Uh so uh historically we worked a lot with the linear channel type of offering to workshop customers, uh, where we could also offer just in time transcoding. So uh with the capabilities that we got from the VPUs, uh we could add a feature of just in time transcoding so that uh when our customers uh record content, we could just save the highest bitrate profile of an ADR stream. And then for the uh on-demand viewing actually retransform on the fly. Uh but as provisioning became possible instantly, more or less, uh in a cloud platform, uh it opened up the opportunity to look at event-based streaming, which is something that we see a growing demand for. Uh a lot of our existing customers do a lot of event-based transcoding now for sporting bikes and so on, but totally new uh opportunities also uh arose from this. Uh, and one of the opportunities that we found was uh together with Snipe. So Snipe does uh eSports, uh, which is a bit new to me. I'm not an expert, so that's why I have Padu with me today to talk a bit more about how they use our technologies in order to make their event-based streaming more efficient and more easy to manage. So all the right.

Per Nybom, Znipe: 5:37

So yeah, I'm so happy to be here showing you how we at Snipe use VPUs in esports, transforming esports. So Snipe is an esports company. We deliver B2B solutions for the biggest tournament organizers and game publishers. And this is a slide of our products that we can deliver, and today I will focus on the multi-POV one, which is the most related to uh how we utilize VPUs. So let's do a quick poll. So, how many of you have played a game in the last week? Raise your hands. We got some gamers? What about in the last month? Sure. What about who has never played a game who's not a gamer? We have some few game non-gamers. Then I will explain these. So these are game titles that we've broadcast with uh the publishers on top or the event organizers. And I will focus mainly on this one game, Apex Legends, because I think it's the best showcase of the strength of VPUs in esports. So for the non-gamers over there, I will uh give you a quick rundown how Apex Legends works. So it's a Battle Royale, which basically means it's Hunger Games the game, or the original Battle Royale movie, if you've seen that one. So it works like each team drop into a map where they run around, gather weapons and supplies, and then they do battle royally until there's one team left standing. Usually the pros play in Teams of 3 with 20 teams in total, and since we stream one POV per team, that's a lot of streams that we need to deliver. So I've recorded a short clip of our product that I will like you to watch with me. So on the bottom here, here, you can see the team logos. And if I click one of these, I can select to view one of these streams. So I pull up the uh stream from that team's point of view. And then on the left here, I have the original event feed, which is a broadcast that is uh tailored and uh curated by the event organizer with a casting crew and uh observers or like a cameraman controlling the camera. And let's play the clip. So you can't really hear it, but you'll see in a second. I I removed the event stream, so now it's just uh one team against the other team. And these are the last two teams standing in the finals. So you can imagine the stakes are high. And it's a shame we can't hear the uh team comms, they're pretty funny. And then we have our winner, so the left team won. Maybe it's hard to tell from Apex, it's a bit chaotic. So to deliver this experience, uh syncing the streams is absolutely critical to get the experience right. So for this, we use a multi-process uh using first AI and image processing, where we pick out a um point of interest in the video feed. So we look at the frames themselves to pick out, usually some sort of in-game clock to synchronize to. And then we also have our uh production crew uh verifying that they are in sync and that they stay in sync. And then for we also ingest uh in-game statistics such as damage events and kill events to um render and generate to render uh like in-game stats tabs and uh generate highlights on the fly. So highlight clips of the most impactful events. Next slide, please. So this uh multi-view streaming product enables us to capture moments that were previously uh impossible to capture because there were just not enough cameras, so to speak. So in the clip you saw, we had the last two Steam teams standing playing uh heads up, and without the multi-view, we would only be able to see one perspective. Uh capturing both enables us to get full immersion in what the teams are feeling. Next slide, please. It also allows super fans, so die-hard fans that only care about one uh team or player to view what they find interesting. Next one and sometimes during these chaotic events, uh the cameraman, so to speak, can't capture all of the uh all of the interesting bits. So he might accidentally flip away from the stream when you were most interested. But with this, you can have your full control of your viewing experience. Next one, please. And this is what all of these contributes to why it's such a loved feature in the Apex community. Next one, please. But it comes at a cost. So Twitch and YouTube, when they broadcast these esports events, they uh stream and transcode one stream, but in our solution we uh transcode and distribute 20 plus streams. So one for each team and the extra event stream. Next one please. And this is where ScaleStream and the Net VPUs come in. So with the help of ScaleStream, we were able to drive down the cost for normal transcoding by 25% and like improve startup times by 70%, which is crucial for event-based broadcasts, where uh we want to minimize the idle time of transcoders. And also working with a team or a company that is more similar size to us uh enables a uh better sense of uh uh personal connection as opposed to the big giant where it mostly feels like you're a customer to a provider. And then maybe the biggest one of them all and the most related to the VPU tech we're demoing here, is that we were able to drive down the cost for high quality transcoding by a mind-blowing 75%. And since esports viewers and other digitally native customers have increasingly high demands for both higher resolutions, higher frame rates, um, we were unable, it was impossible for us to deliver this without the VPUs. It would just be too expensive. But with the help of SKS stream and a 75% increase of efficiency, it's actually possible to deliver this at a reasonable price. So let me go through a story of um when I was uh uh the engineering on call. Next slide, please. So here I am, sleeping soundly in my bed, being the engineering on call, when suddenly I get paged. And it's hello, it's me, it's the production lead. We're going live in 30 minutes, and everything's broken. So I scramble to troubleshoot, pouring over health checks and dashboards and pouring through logs. I'm able to find leave a clue here. Next one. So insufficient capacity. Through all these logs, I found this the culprit. As it turns out, the same instances or it's the same machines that we wanted to use for uh video transcoding is also perfect for training AIs. So it turns out that the uh hours for our uh esports broadcasts really aligned well with the time that AI companies wanted to train. So we would have thought that gamers and AIs share the same work schedule. And I guess uh the moral of this story is that uh with the the help of VPUs and uh avoiding the highly in-demand GPU instances, we hopefully will avoid this issue where I guess until Pontus and the SkateStream guys are able to figure out how to train AIs with these VPUs. So yeah. Next one please. So to summarize, with VPUs we're able to deliver higher resolutions and higher frame rates. This suits very well with the event-based where the startup time is crucial. And just in time workflows to handle the big VOD library that we produce at a reasonable cost. So perfect. And remember, anything is per uh possible with national Swedish ingenuity and passion. Do we have any questions? Perfect. Then thank you so much. And you can find us and talk to us in the ScaleStream booth at this address.

Pontus Eklöf, Scalstrm: 16:59

Thank you so much. Thank you very much. And I might also mention that we uh have a pod over here. So if you uh we can stay around for a while and uh there's a scale stream pod over here where you can have a look at how this works in practice. So please stop by if you have any Matt wave uh go over to Matt and uh you can get a quick demonstration of how this works in practice. So thank you very much. Fantastic.

Voices of Video: 17:26

Thank you.com