How Are Sports Highlights Made? (Manual vs AI)

Sports highlights are made by identifying the key moments of a match, clipping them with accurate in and out points, adding context such as replays, graphics and commentary, and packaging the result for broadcast, video-on-demand or social. Traditionally an operator does this live or immediately after the final whistle. Increasingly, an AI event-detection system does the first pass and a producer refines it.

Both approaches answer the same four questions: what happened, which moments matter, how should they be cut, and where do they need to go. The difference is who — or what — does the work, and how long it takes.

The manual way: inside the truck

In a traditional setup, a logger marks moments as the match unfolds and an editor builds the package against the clock. Someone calls the replays, someone adds the graphics, someone checks the rights and the running order. It is skilled, fast work — and it is exactly where time leaks, because so much of it is sequential and manual. We walked through a real match day in What Actually Happens Inside a Broadcast Truck on Match Day.

The AI way: detect, rank, assemble

An AI pipeline watches the feed, recognises events, ranks them for relevance and cuts clips automatically, so the candidate package exists almost as soon as the whistle blows. The full mechanics are in our pillar guide, How Does AI Create Automatic Sports Highlights?. The editor then starts from a draft instead of a blank timeline.

Manual vs AI at a glance

• Speed: manual is minutes-to-hours after the whistle; AI produces a first cut in minutes.
• Consistency: AI applies the same rules every match; humans vary under fatigue and pressure.
• Editorial nuance: humans still win on context, story arc and rights judgement.
• Scale: AI changes the maths when you cover many matches a week rather than one showpiece.
• Cost: manual cost scales with every fixture; AI front-loads cost into the system, then scales cheaply.

The steps in detail

1. Logging: marking every candidate moment, live or on review.
2. Selection: deciding which moments tell the story of the match.
3. Editing: cutting clips, adding replays, lower-thirds, audio and branding.
4. Review: an editorial and rights check before anything goes out.
5. Delivery: exporting and reformatting for each destination and aspect ratio.

Where AI still needs a human

Rights restrictions, sponsor obligations, tone and the shape of the story are still human calls. An algorithm does not know that a manager’s 200th game deserves a mention, or that a controversial decision needs careful handling. The realistic model is AI for the repetitive detection and cutting, and a producer for the judgement that gives a package meaning.

A short history of how highlights were made

For most of broadcast history, highlights were a craft of tape and memory. An operator marked moments on a logging sheet, an editor cut them on a linear or non-linear system, and the package went out hours later. Faster turnarounds came from better tools and more people, not from changing the fundamental approach. The arrival of computer vision is the first genuine shift in decades, because it changes who finds the moments rather than just how quickly they are edited. Understanding that history explains why the gains from automation feel so large: the bottleneck being removed is one the industry simply learned to live with.

What good highlights actually require

A good highlights package is not just a list of goals. It has rhythm, it respects the story of the match, and it lands the big moments with enough build-up and reaction to feel satisfying. That is why detection alone never produces great highlights — it finds the events but not the narrative. Whether made by hand or by AI, the difference between a competent package and a memorable one is editorial judgement: knowing which moments to dwell on, which to cut, and how to order them so the package feels like the match rather than a spreadsheet of events.