The Opportunity
Could a casual foosball table be turned into a live, data‑driven experience that feels like watching a professional match? The goal wasn’t to solve a traditional business problem, but to compress many of the challenges of real‑time computer vision, analytics, and generative commentary into a single, tangible demo that anyone could understand and enjoy.
A familiar game, untapped potential
Foosball is simple on the surface, but its speed and unpredictability make it a perfect testbed for real‑time computer vision: fast ball movement, frequent occlusions, and constant changes in possession.
Unstructured game data
A foosball match produced no structured data - no stats, no tracking, no history. Turning a casual game into a stream of events, metrics, and probabilities required building a full pipeline from raw video to structured insights.
Engaging the audience
The goal wasn’t just to analyse the game, but to make the experience fun, interactive, and accessible. Any solution had to entertain players and spectators, not just demonstrate technical capability.
The Solution
A vertical camera is mounted above the foosball table, capturing the entire field of play. From this feed, a computer vision pipeline tracks the ball and rods in real time, detecting events such as goals, passes, shot speed, ball possession, off‑target shots, and touches. On top of this event stream, a probability model estimates which team is most likely to win at any moment in the match.
All of this is surfaced through a live dashboard that displays dynamic graphs and real‑time metrics for players and spectators. Each match becomes a data‑driven experience: people can see how possession shifts, how shot patterns evolve, and how the win probability changes after each play.
To make the experience even more engaging, a generative model is used to produce live commentary, giving every game the feel of a sports broadcast. The commentary reacts to events on the table, turning a casual match into something that feels closer to a televised game.
ChutAI premiered at Portugal Google Cloud Day 2024 and has since been used at company events and tech fairs as a showcase for applied computer vision and real‑time analytics. It is designed to be portable and repeatable: the same setup can be deployed at different venues with minimal configuration.
The Impact
ChutAI quickly became the highlight of every event where it was showcased. Players were surprised to see their moves translated into instant metrics, while spectators gathered around the screen, cheering at live stats and laughing at the AI‑generated commentary.
By analysing every shot, predicting win probability in real time, and delivering engaging commentary, the system brought the excitement of professional sports analytics to a simple foosball table.
The technology behind ChutAI extends far beyond foosball. In entertainment, similar setups can power real‑time sports metrics, audience‑engagement tools, and virtual commentators for other sports or events. In industry, the same computer‑vision foundations can support production optimisation, quality assurance, and smarter logistics. The use case serves as a concrete, memorable example of how computer vision and real‑time analytics can move from slideware to something people can see, touch, and play with.
See it in action
