Recap, in ninety seconds.

The whistle goes. The scoreboard locks. And somewhere in the production stack, a recap has already started building itself.

For a broadcast team covering live sport, the window between the final moment and the moment an audience expects a packaged summary is measured in minutes, not hours. The clips editor scrubs. The producer calls time. The graphics person is already two screens ahead. Everything converges on the same deliverable: a watchable summary of the event, start-to-finish, branded, captioned, ready.

Recap automates the assembly that precedes that deadline. Given a completed event in the archive, it identifies the significant moments — the goals, the wickets, the turning points — sequences them with transitions, applies the brand layer, and outputs a finished video without requiring an editor to open a timeline. The whole process takes less time than the post-match interview.

01What the recap window actually costs

The cost of post-event recap production is rarely calculated honestly. The direct cost is headcount: editors working odd hours to cover live windows, assistants pulling clips from shared storage under pressure, graphics teams pasting brand assets into a timeline by hand. The indirect cost is harder to see — the recap that ships twenty minutes late, after the audience has already moved on. The moment that was almost in the package but got dropped because there was no time to scrub for it.

For a broadcaster covering a regional sport, those costs are structural. The game schedule is not negotiable. The audience window is not negotiable. The edit team is sized for the average match, not the final. Recap does not eliminate that pressure. It changes where the pressure lands. The editor's job becomes review and judgment, not assembly and timing.

The edit that mattered was never the assembly. It was the call about which moment the recap led with.

That distinction — assembly versus judgment — is where automated recap starts to reshape what a production workflow looks like. An editor who is not scrubbing for clips is an editor who is watching the output and deciding if it is right. That is the job that benefits from a trained eye. The system handles the part that scales. The editor handles the part that cannot.

02How the output arrives

The recap output is a single MP4: burned-in captions, intro and outro slates pulled from the organisation's brand kit, clean transitions between moments. Nothing about it requires a non-linear editor to open. The editor clicks play and decides whether the cut is what they would have made. Most of the time it is close enough to be the starting point. Sometimes the call is to pull one moment and replace it with another — which takes a second query, not a second scrub.

The clips that feed a recap come from the same retrieval surface that answers a search query. If an editor wants to know whether the system found the best moment from the first half, they can ask. The clips are citable, timestamped, and traceable to the source file. Nothing in the output is invented, and nothing in the output is unfalsifiable.