For the Milano Cortina Games, Olympic Broadcasting Services (OBS) is delivering more than 6,500 hours of content, with more than 900 hours of live action, spread across eight sports, 16 disciplines, and 116 medal events.

To aid it with the Games-on-Games uplift in output, artificial intelligence (AI) is being used to assist with managing the volumes of content.

Building on the successes of Beijing 2022 and Paris 2024, OBS is deploying and trialling a suite of tools designed to streamline production workflows, shorten content discovery time, and deliver personalisation. OBS said that what began as “experimental” technology has now become a strategic pillar of OBS’s broadcast innovation.

As well as being used for replays and analysis, OBS has also embraced AI to generate highlights in real time and provide instant transcription and translation, which OBS said redefined broadcast workflows, delivered speed, scalability, and technical excellence for the next generation of Olympic coverage.

Two solutions in particular – Automatic Media Description and AI Highlights Generation – are designed to optimise workflows, enhance creativity, and enrich storytelling across platforms.

AI Highlights Generation

This AI-powered service enables OBS and media rights holders (MRHs) to swiftly create tailored highlights, using advanced triggers, ready for multi-platform distribution. By automating workflows and reducing search time and manual editing, it accelerates personalisation and delivers “high-impact” moments that boost audience engagement.

First introduced at Paris 2024, OBS’s AI Highlights Generation service was adopted by 14 broadcast organisations, producing more than 100,000 highlight clips. Swimming generated the most automatic highlights with approximately 13,900 clips, closely followed by athletics with 13,600 clips. At Milano Cortina 2026, MRHs can create custom compilations, such as top plays, athlete profiles, and social media-ready content, from all 16 disciplines.

Automatic Media Description

Currently in its early development phase, Automatic Media Description (AMD) uses visual and large language models, audio signal processing, and structured data integration to generate real-time, metadata-rich descriptions of broadcast content. Designed to enhance indexing and retrieval, AMD applies computer vision for scene recognition, synchronises commentary audio for contextual tagging, and incorporates event data for semantic structuring.

While currently supporting internal workflows, AMD is engineered to scale into integrated services that enable broadcasters to implement semantic search, natural language query interfaces, and generative content pipelines, optimising production efficiency and accelerating content discovery across multi-platform environments.

Automated Multi Clips Feed Logging

OBS Engineering is collaborating with the Archive team on a proof-of-concept initiative to automate video logging for Multi Clips Feeds (MCFs) — secondary feeds offering unique angles of the competition. Driven by the Archives team, the project seeks to enhance and scale MCF logging through AI-generated descriptions and advanced content analysis, addressing the growing volume and complexity of this material.

Archive metadata quality control specialists will review and validate the AI outputs to ensure accuracy. If the approach proves effective, it could be gradually deployed in production, streamlining MCF logging workflows and delivering improved services to OBS and MRH production teams.

OBS chief technology officer Sotiris Salamouris said: “AI is here now – especially with the latest developments providing us with additional capabilities in this space, which will allow us not necessarily to replace our live loggers, but to extend logging to a lot of other content that is very lightly logged because of the volume that we have and the complexity of doing this.”