Capturing Athletes for AR: Inside a Player Capture Session

Why the capture session is the project

When people discuss an AR fan experience they discuss the screen. In our experience the screen is the easy half. For the Ice Hockey World Championship 2024 the hardest part was not the signage but the source material: each hockey legend had to be captured in a studio session and processed into an AR-ready figure that still reads as that specific player at full scale.

The reason is a comparison the viewer makes automatically. An AR figure standing beside a fan is judged against the person next to it. Proportions, lighting and how the figure meets the ground in the frame are assessed against a real body a metre away, which is a far harsher test than any product render faces. A hand-modelled 3D player looks wrong at that scale, which is why we photographed the players themselves and worked from that.

Who was in the room

The session featured Jaroslav Špaček, Martin Straka, Radko Gudas, Petr Nedvěd, Ondřej Pavelec, Tomáš Plekanec and Jiří Šlégr.

That line-up was not assembled only for name recognition. It spans defencemen, forwards and a goaltender, and the difference matters once the figures are in the scene: each position brings a different stance, a different silhouette and a different set of gear. Fans read those silhouettes instantly — a goaltender's outline is not a forward's outline — so covering the positions gives the installation variety that the audience actually perceives.

What was captured, and from where

Each player was photographed in full kit under controlled lighting, from the angles the AR scene would later need. Working backwards from the scene is the whole method: you do not photograph a player and then decide how to use the images, because the compositing conditions dictate what is usable.

The mismatches that break the illusion are small and specific. The direction of the key light, because a figure lit from the wrong side next to a fan under mall lighting reads as a cut-out. The height of the camera relative to the player, because that sets the eye line and the perspective the composite has to match. And how a stick or a glove sits at the edge of the silhouette, because equipment produces exactly the kind of thin, complex outline that betrays a poor edge.

We shot multiple takes per pose. Players are not models, and the poses that read best on screen are usually not the posed ones — the celebration mid-turn or the relaxed stance between shifts hold up better next to a real person than a static portrait.

From photographs to real-time AR figures

From the capture material we produced AR-ready figures that the application composites over the live camera feed in real time, so the fan sees themselves and the player together on the screen as they move.

Two constraints shaped that processing work. The composite had to render fast enough to feel like a mirror rather than a delayed video — the illusion depends on the figure being present in the same instant as the fan's own movement. And it had to be legible at large-format signage size, where compression artefacts and edge halos that are invisible on a phone become obvious.

The interaction on top of all this is deliberately thin: no app to install, no menu to learn. A fan steps into frame, sees the player appear beside them, poses, and the photo is captured. All of the complexity is spent on the figure so that none of it has to be spent by the visitor.

The awkward parts

Turning an athlete into an interactive element has friction that a product shoot does not.

Studio time with professional players is scarce and not extendable. Whatever the storyboard did not anticipate is not obtainable a week later.

Gear complicates the edge. Sticks, gloves and pads generate the outlines that are hardest to keep clean, and they are precisely the elements fans use to identify the player.

Athletes hold sporting poses, not photographic ones. The useful frames come from movement, which means shooting more and selecting later.

The target size is unforgiving. Material judged on a studio monitor is a poor proxy for a figure three metres tall standing next to a person.

From studio to the street in Prague

The deployment itself is the simple half — and it is simple because of the capture work. Large-format digital signage stood at locations across Prague during the Championship, each screen running the TapGest AR application, compositing the processed player over the live camera image in real time so fans could take a photo or a short video with them.

The signage became a focal point during the tournament, with fans queueing to be photographed next to a legend, and it gave the organisers a branded activation that could be placed at several locations across the city rather than in a single fan zone. That multiplication is only affordable because the expensive part — the figures — is produced once and reused at every screen.

What we took into later projects

Two lessons carried forward from this session.

Capture more poses than the storyboard calls for. Once the scene is running in a real space, the pose that works best is often one nobody planned, and studio access is not repeatable.

Test at final size and in the actual lighting of the location as early as possible. A scene that is convincing on a studio monitor can fall apart under mall lighting on a screen three metres wide.

A third point is a practical consideration rather than a documented outcome: when scoping a fan activation, the binding constraint is usually not the technology but who you can realistically get in front of a camera and for how long. That answer should shape the concept before anything is built.

Related work

The same pipeline — capture real people, composite them live, keep the interaction to one step — sits behind our AR digital signage and AR photo booth work, including the Fortuna League fan activation and the Kafka photo experience at Quadrio Mall. The sector view is on our sports and entertainment page, and how these formats differ on site is covered separately.