Interviews

AI Cameras Are Just the Beginning. Matúš Bubeník: CBHTV Should Become the New Home of the Ball Hockey Community

Automation and artificial intelligence are gradually changing the way fans can watch ball hockey games. The new CBHTV platform brings broadcasts from competitions across different levels and, in the future, is expected to offer additional content, analytical tools and easier ways to work with video.

The technological solution is provided by Perinvest Media, whose cooperation with Czech ball hockey began at the 2023 Junior World Championship in Liberec. Matúš Bubeník, CEO of Perinvest Media s.r.o., answered questions from hokejbal.cz.

The cooperation began at the Junior World Championship

Why did Perinvest decide to enter the world of ball hockey? What attracted you to the sport?

Our first contact came in 2023, when the federation approached us to produce the Junior World Championship in Liberec. From the perspective of the streaming service we delivered, the event was very successful both technically and economically, and it helped launch the idea of cooperation at league level as well.

In what ways do you think streaming and modern technology can give ball hockey an interesting future?

A huge amount of content can be produced that will serve the sport in many different ways. For example, it can be used for coach and referee education, disciplinary committee work and much more. At the same time, the content will be available not only to dedicated fans who attend games in person, but also to family members who, for whatever reason, are unable to travel to matches.

It will also reduce the workload for people currently involved in manually producing only selected games. Our cameras can record from morning until evening across all age categories without increasing costs or placing additional demands on clubs.

How does an AI camera actually work?

The term “AI camera” sounds simple, but what is actually behind it? How is such a broadcast produced?

Basically, it is a simple system, but there are many complexities behind it. The camera is connected to a server running software with an AI algorithm. That algorithm controls the camera, stores the footage and sends it through distribution networks to the end user.

Everything is linked to the information system and the match schedule, so the camera knows when a game is taking place and which teams are playing. It starts automatically, records the game and then switches itself off.

How does the camera know where to look? Does it follow the ball, the players or the overall flow of the game?

It primarily follows the game itself, but every match that is recorded provides more data for further training and fine-tuning. AI is improving at an incredible pace today and learns very quickly, but it still needs inputs. That means new venues, new situations and new scenarios.

In ball hockey, we are not yet at a 100% perfect level, but the results we are already able to deliver are very good and will only continue to improve.

Is ball hockey a difficult sport for artificial intelligence because of the small ball, quick changes of direction, boards and player battles? Is special software being developed specifically for ball hockey?

As I mentioned, ball hockey is a smaller sport, so the AI is not yet perfect. The number of recorded hours is certainly lower than in ice hockey or some other sports, but the algorithm itself is very similar.

The type of ball, rink dimensions, changes of direction and similar elements are simply variables within the entire system, which is built on the same foundations. Today, we work individually with ice hockey, ball hockey, floorball, field hockey, inline hockey and other sports.

Technology keeps advancing. One camera is not the final solution

What are the biggest limitations of automated cameras today? What can AI still not do as well as a human camera operator?

We have already come a long way. Of course, we cannot replace a person physically present at the venue—if someone kicks out a cable, a person can simply plug it back in and continue.

But when it comes to broadcast quality, remote productions controlled from a central facility, multi-camera AI broadcasts and other solutions already exist today. As always with technology, it comes down to cost. Over time, costs will go down while quality will continue to improve.

What happens if the internet connection drops during a game or another technical issue occurs?

That is obviously a problem. The broadcast runs through the network, and without an internet connection the signal cannot be transmitted. But the same applies to manual production.

Everything is controlled remotely, so if the equipment is disconnected from the internet, we cannot access it remotely to fix any potential issue.

Is today’s single automated camera the final solution, or could we eventually see multiple AI cameras, automatic replays and something close to full television production?

It is definitely not the final solution. Quite the opposite—it is only the beginning. More advanced options already exist, but naturally they are significantly more expensive.

CBHTV as the new home of the ball hockey community

CBHTV is not only about live broadcasts. What else should the platform offer fans, and where do you want to take it technologically in the future?

Exactly. In a way, it should become the new home of the ball hockey community. A place where people can quickly and easily find live broadcasts and recordings from all age categories.

They should be able to find their favorite players and gain quick access to clips involving them. We are planning player accounts, while the federation will have a place to showcase its partners, and clubs will have the same opportunity.

The solution also includes analytical tools available to clubs and the federation, such as team accounts, media accounts and referee accounts, making work with content easier and more efficient.

Data is a major topic. Could AI eventually create goals, highlights, clips of individual players or even social media content automatically?

Absolutely. A certain form of this already exists today and, as I keep saying, it will only continue to improve.

The broadcast of the future? A combination of AI and manual production

CBHTV will become a paid service from October. What exactly does a fan finance through their subscription, and why is a paid model important for the future development of the project?

That is a very good and important question. People need to understand that everything mentioned above costs money—and not a small amount.

As a partner of the federation, we take 100% of those costs and the associated risk upon ourselves. We provide a fully managed service. In this case, the federation supported the cooperation and financed the hardware, but purchasing the equipment is only the entry point into the whole project and a one-time expense.

There are then recurring costs for software, AI licences, storage, distribution, the OTT platform, technical and customer support, further development, administration and management. All of these form a chain of monthly expenses without which the service could not operate, develop and provide the sport with the content we have been discussing.

All of this is financed through subscriptions. If there are not enough subscribers to cover the costs, then we will have made the wrong bet. But we are willing to take that risk because we believe we are bringing a valuable product into the ball hockey environment.

In modern society, most services are already behind a paywall. We have been developing this model in Czechia since 2021, when we first introduced paid ice hockey broadcasts. We are now in the sixth year of that cooperation, and clubs and fans are constantly asking us to add more content and cover more venues.

Looking three to five years ahead, what could the ideal ball hockey broadcast of the future look like?

I would not dare predict how quickly this field will develop, but a combination of AI and manual production, multi-camera AI broadcasts, remote game production controlled by a single operator instead of sending three- or four-person crews to venues, automatic highlights and similar tools are already available today.

In a few years, they will simply be even more accessible.