How to Cut Football Highlights From a 2-Hour Match

A method for long match recordings, covering how many candidates a two-hour file produces, how to review them quickly and how to export large files safely.

To cut highlights from a long match, AutoHighlight proposes candidate clips by looking for the loud moments in the recording's audio, capped at roughly one per two minutes of footage. What you get back is a review list, not a finished reel, and at two hours that distinction starts to matter.

A two-hour recording is a different problem from a twenty-minute one, and not only because it is longer. At that length the review itself becomes the bottleneck, the file gets big enough to hit real technical ceilings, and a careless export setting can cost you a whole evening.

Here is a method that holds up at that scale.

Know the shape of what you are about to get

The automatic loudness pass returns at most one candidate per two minutes of footage. For a two-hour file that is a ceiling of about 60, each around ten seconds long — seven seconds before the loud moment it found and three seconds after, clamped at the ends of the recording. Only peaks that clear the recording's own average count, so what decides the number is contrast rather than volume: a match where nothing stands out returns few or none, however loud it is overall.

At a full pool that is ten minutes of review, if you spend ten seconds per candidate. That number is the whole design: it is a recall pool sized to be thrown away fast, not a list of 60 things that are each worth keeping. Going in expecting a finished reel is the quickest way for this to feel wrong.

A useful way to think about it: the audio pass answers "where did anything happen", and you answer "was it worth keeping". Those are genuinely different questions, and only one of them can be automated on footage like this.

Review in two passes, not one

On a long match, reviewing linearly and perfectly is a trap — you spend real care on candidate 9 and then rush candidates 45 through 60 because you are tired. Split it:

Pass one — keep or kill, no trimming. Play each candidate, decide within a couple of seconds, move on. Do not adjust in and out points yet. Expect to throw most of a full pool away — it is sized for recall, not precision — and treat whatever survives as the real shortlist. We have no measurement of how many that usually is; it will depend on the match.

Pass two — tag and trim the survivors. Now the careful work happens on a small set. Tag each one Goal, Shot, Save, Defense, Pass, Skill or Other, then put it into edit mode with its Edit control and drag the in and out handles until it opens on the move rather than on the reaction.

Doing it in this order means your attention is spent where it changes the output. It also means that if you are interrupted, you stop with a usable shortlist instead of half a match reviewed to perfection.

Why some real moments will be missing

Worth knowing before you go looking for a setting that fixes it: the pass finds loud moments, and among peaks less than 18 seconds apart it keeps only the loudest one. The quieter peaks near it are dropped, not merged. That is what stops a goal, its cheer and the restart from becoming three entries — but it also means that when two genuinely separate things happen inside that window, only the louder one becomes a candidate, and the other is simply not in the list.

Quiet moments are the bigger gap. A tidy passage of play that nobody shouted at will not appear, because nothing in the audio marks it. On recordings with no touchline noise at all, the pass has very little to work with and manual marking is the honest answer.

The file-size rules that actually bite

Long recordings are large recordings, and there are three thresholds worth holding in your head:

• 3 hours and 35GB are the hard limits per file. The size is checked when you add the file and the duration as soon as it is known, so an oversized recording is refused up front rather than failing halfway through.
• More than 20 clips in one export raises a confirmation first. On a two-hour match it is genuinely easy to still have thirty-odd clips selected from pass one, and each one becomes its own file.
• Disk space at the other end. Fifteen ten-second clips are small, but if you keep sixty of them from a high-bitrate source, check where they are landing.

Usually not a limit: a batch export works through the clips one at a time and reads each from the original file, so a 12GB or 20GB match does not normally need to be held in memory. That streaming route depends on a browser feature; where it is unavailable the export falls back to loading the whole source into memory, and a multi-GB file can fail there. Exporting a few clips at a time is the workaround.

What happens when you press export

The cut runs in your browser. When a clip needs no overlay, no rotation and no filter, it is copied straight out of the source without re-encoding, which is fast and barely depends on how large the source file is. When a re-encode is needed, it is scoped to the segment rather than the match — so the cost tracks the clip's ten seconds, not the file's two hours.

One consequence worth knowing: a copied cut cannot start mid-keyframe, so it begins at the keyframe at or before your in point. Keyframes sit a few seconds apart in most recordings, which means a fast copied export can open a second or two earlier than where you dragged. The end lands where you asked. When a clip does need a re-encode, the start is placed exactly. In practice this is a run-up you did not ask for rather than a missing moment, but it is why trimming to the exact frame is not worth your time on a copied export.

Export speed depends on your own machine, since the work happens there rather than in a queue on a server.

A realistic time budget

For a two-hour recording, from opening the file to having clips on disk:

• Adding the file and running the audio pass: one pass over the audio track, scaling with the length of the recording.
• Pass one over a full pool of ~60 candidates: about 10 minutes at ten seconds each.
• Pass two, tagging and trimming the survivors: 10 to 15 minutes.
• Exporting: minutes, and mostly unattended — though some browsers, iOS in particular, hold a multi-clip batch until you come back and save the files.

That arithmetic is the shape of the work, not a measured figure — it assumes a full candidate pool and ten seconds of attention per clip, and your own numbers will differ. Even doubled it compares well against scrubbing the timeline, on an evening when you have already spent two hours at the pitch.

Adding a match and previewing the candidates does not need an account; exporting the clips it found does, and the editing flow is free to use today (pricing). The step-by-step version of this flow is in how to make football highlights from a full match, and the settings that make a long recording easier to work with are in best format and settings for football match footage.