How ReCAP detects and logs on-screen timers and countdowns

The ReCAP initiative, funded through the European Union's research framework, has spent the past several years pushing the boundaries of what machines can recognise inside a broadcast video stream. Its toolset already spans face recognition, logo detection, quality monitoring and duplicate-content identification, all running in real time against live feeds. One capability that often goes unnoticed sits closer to the edges of the frame: the small graphics that tell viewers how long until a programme begins, how much time remains in a sports segment, or how many seconds a presenter has left to wrap up.

Detecting those visual cues matters more than it first appears. In Australia, broadcasters such as the ABC, SBS and the commercial networks routinely use on-screen countdowns during live crosses, election coverage and sporting broadcasts. When a network's automation or compliance team needs to verify that an advertisement ran for the right duration, or that a segment ended exactly as scheduled, having a machine-readable record of every visible timer removes hours of manual scrubbing. ReCAP treats these graphics as first-class metadata, just as it does faces and channel logos.

Why on-screen timers carry operational weight

Timers and countdowns function as a contract between the broadcaster and the audience. They promise that a news bulletin will resume at a precise moment, that a cricket over will end at a known boundary, or that an advertisement will give way to the next programme without delay. From a production standpoint, those promises must be honoured frame by frame. A post-match review at Fox Studios in Sydney, or a compliance audit at the regulator's Canberra office, often hinges on whether the on-screen clock behaved as scripted.

The same logic applies to live events streamed from stadiums such as the MCG, ANZ Stadium or Adelaide Oval. Operators frequently overlay temporary graphics — a "3 minutes to kick-off" banner, a "starting soon" plaque — that exist only for a few minutes and never appear in any recorded schedule. ReCAP's ability to spot and log these ephemeral elements gives broadcasters a far richer archive than a human logging team could ever assemble by hand.

How the system recognises graphical time displays

Timer graphics have a recognisable visual grammar. They tend to sit in the lower third or upper right of the frame, use high-contrast colours against the picture, and contain digits that change at predictable intervals. ReCAP's analysis pipeline samples incoming frames at high frequency and runs a dedicated detector over candidate regions. The detector checks for digit-like glyphs, the presence of separator characters, and the typical aspect ratio of a two- or four-digit display.

Crucially, the system is not searching for a single fixed template. Broadcast designers in Australia and elsewhere regularly refresh their on-air look — AFL broadcasts on Channel Seven have changed their clock styling more than once in recent seasons, and SBS uses a noticeably different palette during news bulletins. ReCAP therefore uses a combination of contour analysis, optical character recognition tuned for short numeric strings, and temporal reasoning to confirm that what it has found is actually counting down rather than showing a static channel identifier.

From detection to frame-accurate metadata

Once a timer or countdown is identified, ReCAP attaches a structured record to the video timeline. Each occurrence carries the exact frame number, the displayed value at key sample points, the bounding box of the graphic, and a confidence score. This metadata flows into the project's central processing layer, where it can be aligned with other detections — for example, the appearance of a particular presenter, or the black frame detection work that ReCAP has published separately for monitoring signal integrity.

Because the logging happens in real time, a producer in Brisbane can query the live stream and ask whether a countdown has been on screen continuously for the past ninety seconds. The same query, applied to a recorded asset in a media asset management system, returns the full historical record of every timer that appeared during the programme, ready for compliance reporting or editorial reuse.

Live sports, news and election coverage in Australia

Australian broadcasting schedules are dense and varied, which makes timer detection unusually valuable. Cricket broadcasts on Foxtel and the free-to-air networks include over-by-over clocks, drinks-break countdowns and innings-end timers. Rugby league coverage features shot clocks and sin-bin countdowns, while AFL matches include the central game clock and the supplementary thirty-second clock used during stoppages.

News and current affairs bring a different set of patterns. ABC News Sydney's late bulletin uses a "next in 60 seconds" graphic before commercial breaks, while SBS World News often shows a countdown to a live cross from Canberra. Election nights, broadcast jointly by several networks from studios in Sydney and Melbourne, layer multiple countdowns over each other as results are tallied. ReCAP logs each of these independently, even when they overlap or change position during a broadcast.

Edge cases the detector must handle

Real broadcasts are messier than test footage. Timers can be partially occluded by a lower-third strap, fade in and out rather than appear suddenly, or be rendered in a font the system has never seen before. Some countdowns tick backwards while others, such as elapsed-time clocks during a long interview, tick upwards. ReCAP handles these variations by combining visual heuristics with a small temporal model that expects the digits to change every second.

Animated graphics present another challenge. A countdown might dissolve, scale or slide between updates, briefly disappearing from the frame before returning in a new position. The detector treats brief absences within an expected window as the same graphic, splitting the record only when the timer vanishes for several seconds or a fundamentally different display takes its place. This keeps the metadata clean without over-segmenting every stylistic transition.

Logging, storage and downstream consumption

Detected timers and countdowns are stored in the same metadata schema ReCAP uses for its other detections, which makes them searchable alongside faces, logos and quality events. A typical query against an hour of Australian prime-time television might return dozens of timer records, each tagged with the programme name, the channel, the time of appearance and the displayed values at regular intervals.

That record is useful long after the live broadcast ends. A media asset management team can reconstruct exactly when a presenter was counting down to a commercial break, when a sports clock reached zero, or when an election graphic signalled a state-by-state result. Compliance teams can verify that an advertisement was accompanied by the correct countdown, and editorial teams can quickly locate every clip in which a particular timer style was used.

Scaling detection across multiple channels

Australian broadcasters often run many simultaneous feeds: a main network, a multichannel companion, regional variants for New South Wales and Queensland, and on-demand streams. ReCAP's detection pipeline is designed to run in parallel across these feeds without contention, sharing model weights but maintaining independent logging channels.

The practical benefit shows up during events such as the Melbourne Cup, when multiple networks broadcast the same race with different on-screen graphics. ReCAP can process each stream in real time, log every countdown unique to that feed, and deliver a consolidated view to a control room in Sydney without missing a frame. The same infrastructure scales gracefully to archive processing, where the system can chew through weeks of recorded content overnight and surface every timer that ever appeared.

Practical steps to integrate timer detection into a broadcast workflow

• Connect the ReCAP analysis service to the live feed ingest points used for main and secondary channels, including any regional variants such as those serving Perth or Darwin audiences.
• Map the timer's bounding box, displayed value and confidence score into the existing MAM schema so editors and compliance staff can search for timer events without learning a new interface.
• Define alert thresholds for unusual timer behaviour, such as a countdown that stalls unexpectedly or a clock that appears outside the scheduled programming block.
• Retain the full metadata record for at least the regulator-mandated archival period, and longer for high-value content such as state football finals or federal election coverage.
• Periodically review the detector's output against a human-verified sample to confirm that the visual heuristics continue to match the broadcaster's current on-air graphics.

Across news, sport and live events in Australia, on-screen timers and countdowns function as small but reliable signals that production has gone according to plan. ReCAP turns those signals into structured, searchable metadata that arrives the moment the graphic does, and stays useful long after the broadcast has ended. The capability quietly closes a long-standing gap between what viewers see on screen and what the broadcaster can prove happened, leaving production teams with a clearer, frame-accurate record of every second they put on air.