SVG
article here
As part of its drive to generate more content with the same or fewer resources, the Bundesliga conducted a live test of a fully automated pan-tilt-zoom camera system incorporated into the live host feed of the Franz Beckenbauer Supercup.
The live integration has been deployed on other football leagues and others sports before but this is a first for the Bundesliga and also for its broadcast partner Sky.
“We’ve done several tests, including a fully automated tactical feed last week in Hannover for a Bundesliga 2 match,” explained Dave Greene, Sales Director at Dutch AI specialist Studio Automated.
“We’ve already deployed automated PTZ in ice hockey, handball and football in other leagues. This the first time we’re integrating a technical live AI feed into the Bundesliga match experience together with Sky.”
One camera was in operation at Boroussia Dortmund’s Signal Iduna Park positioned on the main camera gantry. This was mounted on a HotHead 3 remote head from Egripment Camera Support, part of the Broadcast Solutions Group (BSG).
“On‑site it’s plug‑and‑play,” Greene said. “The camera can be reached remotely, so in a centralised production someone can simply access the system over the internet and start or stop the AI through our API. You can connect multiple cameras. The remote head behaves like a fluid head on a tripod, fully integrated into the OB van infrastructure.”
With Greene was Egripmont CTO Martin Dresbach. He described the HotHead 3 as a fully digital, IP‑based remote head with an API that Studio Automate uses for control. “It can carry heavy packages up to 60 kilos, so large lenses up to 122x are no problem. We provide the hardware; the magic happens in the Studio Automate software.”
Jens Fischer, senior business development manager for Global VAR and Sports Solutions at BSG explained that Studio Automated supports 15 professional sports out of the box.
“It has trained an AI to control cameras exactly as you’d expect in professional sports Naturally, for the Supercup it is using the AI model for football. It’s a multi‑camera model that supports standard football positions: Camera 1, Camera 2, the 16‑metre tactical camera, left and right offside, and the high‑behind‑goal angle. Today we’re running the AI model for the tactical feed.”
The team were keen to emphasise that the AI is not simply ball or object tracking. It has been trained to understand the game.
“The AI doesn’t play ‘catch the ball’ like you might expect,” Fischer said. “It actually understands football. We didn’t hard‑code rules—the AI learned them on its own.”
A corner‑kick example explains this well. Traditional automated systems track the ball and keep it in the centre of the frame. That produces bad pictures, according to Fischer, “half the audience on one side, the ball buried in the corner.
“A real camera operator would frame the corner flag on the left and bring the ball into the shot,” he added. “The AI learned that behaviour independently.”
The AI is also smart enough to understand why a person in black on the touchline never enters the pitch, or why there suddenly appear to be 23 players—because substitutions have happened. “It understands the logic of football and frames the game the way a human operator would.”
Dresbach stressed that the AI doesn’t manipulate the video. “It receives the low‑latency feed and instructs the remote head how to move. Production companies take the direct feed, match it with other cameras, and shade it in real time.
“It’s important to understand that the AI has been trained on football movements and anticipates play. If you track only the ball, you’re always too late—and you frame incorrectly, especially on corners. The system detects the integrity of the game, not just a single object.”
Asked how accurate the system was Greene replied, “We work with top‑tier football leagues that demand extremely high uptime. Football has unique situations, but through iterative updates we can teach the AI how to respond.”
The system control can either be on‑premise (which it was at the Supercup) or at a remote production centre. “The AI analyses the live feed coming from the camera lens and sends instructions to pan, tilt and zoom, Dresbach said. “The remote head then executes those movements, giving you the natural look and feel of a football match.”
While a tactical feed was demonstrated here, the Studio Automated-BSG combination can offer solutions for Camera 1, left and right offside, and high‑behind‑goal positions.
“It gives production companies flexibility to blend manual and automated production,” Greene said. “If you have a traditional production and want to add more cameras cost‑effectively, we provide a way to do that.”
Pressed as to whether this inevitably leads to fewer camera-ops down the line, Greene said:
“Things are changing on the operator side. In five years we may have fewer camera operators, or simply more flexibility. Recently at a EuroLeague playoff—a very downscaled production—they used one AI tactical camera because the budget allowed adding one more camera. In future, automated 16‑metre cameras could let productions stay within budget while increasing coverage. It’s not about replacing people; it’s about doing more with the same budget.”
Supervising this and a whole array of other tech trials was Alexander Günther, Managing Director of Sportcast. The wholly-owned subsidiary of the DFL produces the host signal for all Bundesliga and BL2 matches – some 700 events a year.
“This is part of our methodological approach: idea, concept, POCs, iterative development, and finally delivery. And today—it’s delivery,” he said.
The AI-powered feed will be evaluated but likely rolled out as standard for every Tuesday 20:30 match this season.
At the Supercup, Sportcast were hosting the match not only for the Bundesliga match but also Sky and other domestic and international partners. The AI tactical feed was in addition to the ‘platinum setup’ of 28 cameras which it reserves for the bigger games such as Der Klassiker.
While Sony and GV systems cameras are in dominant use across the season this may change as Sportcast evaluates ARRI live cameras.
“We tested ARRI cameras at Sports Innovation two years ago, and yes, we’ll continue testing them. They offer different shading, different perspectives, and a more cinematic look, which is increasingly in demand for documentaries.”
Volumetric capture was also being tested with twenty cameras arrayed around the pitch perimeter. “It’s non‑live for now, mainly for highlight editing, but it looks very good,” Günther said.
Even Meta glasses were being used for the first time in an early exploration of the potential angles and fresh point of view a live stream from wearers (in the tunnel for example) might bring.
Günther echoed the thoughts of Bundesliga CEO in highlighting the macro challenge that he now faces which is balancing the shared experience of soccer with personalisation.
“Historically there was one world feed. Now we’re exploring how to provide both a world feed and individualised streams.
“Centralised production helps enormously. It allows us to bring all feeds into one hub and distribute them to different editorial teams. Automation and AI will change this further—fewer camera operators and more potential for personalised feeds. The next big innovation will be integrating 3D data and audio into a completely new production layer.
“Our primary responsibility however is to provide the correct source. Audio, video, and data has to be accurate and captured live from the field of play. Fans need to trust that what they see is real.”
No comments:
Post a Comment