How to Make Football Highlights From a Full Match

A step-by-step walkthrough for turning a full 90-minute match recording into a short highlight reel, from adding the file to exporting clips by tag.

AutoHighlight turns a full match recording into short clips in the browser: you add the file, an audio pass proposes the moments worth watching, and you trim and export the ones you keep. This guide walks that route end to end, using a 90-minute recording as the example.

If you record your own matches, the hard part was never the camera. It is the two hours of footage afterwards. Somewhere in there are four goals, a disallowed one, two good saves and a moment your left-back would like framed — and finding them means scrubbing a timeline until your eyes hurt.

This guide walks through the whole route: full match in, short clips out. It covers the loudness pass, which runs on your own device. AutoHighlight also has an optional speech feature that works differently: it sends the extracted audio track — audio only, never the video — to a speech service, and only after you have been asked and agreed. If you use it, the transcript is kept server-side for a limited period. The privacy policy has the specifics. Nothing in this guide requires it.

Before you start

You need one thing: the match file, on the device you are sitting at. A tripod camera, a Veo or similar fixed camera, a parent's phone from the halfway line — any of them can work, provided two things are true: the recording has an audio track, and your browser can actually decode it. Audio is what the first pass uses to find moments, so a camera with the microphone disabled will give you very little to work with.

That second condition is worth spelling out, because the two are checked at different moments. The file picker accepts a file by its type or extension, but the analysis needs the browser itself to decode the video and report a duration. MP4 with H.264 and AAC is the combination browsers handle most reliably; some older or less common ones (AVI, WMV, FLV, or HEVC inside any container) may be accepted at the picker and then fail to play. If that happens, convert to MP4 first.

The app accepts MP4, WebM, MOV, M4V, AVI, MKV, FLV, MPEG, 3GP, WMV and TS video, plus MP3, WAV, OGG, AAC and M4A if you only have audio. A file may be up to 3 hours long and up to 35GB. Anything past either limit is refused before processing starts rather than failing halfway through.

Step 1 — Add the match

Open the app and drop the file onto the upload card. Nothing is re-encoded at this stage; the browser reads the file and gets its duration and audio track so it can start looking for moments.

Step 2 — Let the audio pass propose candidates

This is the part that saves the two hours, so it is worth understanding what it actually does.

The audio track is measured in half-second windows. Each window gets a loudness value, those values are then blended with their neighbours so the measurement is less jumpy, and the app looks for local peaks that stand clearly above the match's own baseline — not above some fixed number, but above the average loudness of your recording plus a margin. A windy park pitch and an indoor hall each get judged on their own terms.

Two more rules shape what comes back:

• Of any two peaks less than 18 seconds apart, only the louder survives. The quieter ones near it are dropped, not merged — which is what stops a goal, its cheer and the restart from becoming three entries.
• Each candidate clip normally runs from 7 seconds before the peak to 3 seconds after it — about ten seconds. Both ends are clamped to the recording, so a candidate right at the start or the end comes back shorter. The anchor is simply the loudness peak; when that peak is a crowd reaction it tends to trail the play that caused it, which is why the window carries far more lead-in than tail.

The number of candidates is capped at roughly one per two minutes of footage — about 45 for a 90-minute match. That is a ceiling, not a promise: only peaks that clear the recording's own average count, so the number is decided by contrast rather than volume — a recording with nothing standing out returns few or none however loud it is overall.

That cap used to be twice as generous, at one per minute. Halving it was a product judgement rather than a measured result: on real match footage the tail of a 90-candidate pool was judged to be mostly peaks that barely cleared the threshold, and cutting it trades a little recall for a list people actually finish reviewing.

Step 3 — Review and tag

The clip list shows every candidate — there is no paging to fight. Click one and it plays from its own start point, so reviewing a candidate costs ten seconds, not a scrub through the timeline.

Each clip can carry a tag: Goal, Shot, Save, Defense, Pass, Skill or Other. Tagging is not decoration — it is what lets you export a goals-only reel for the club feed and a separate save-and-skill reel for the goalkeeper, from one pass over the same match.

Work fast here. Plenty of candidates will be ordinary noisy moments — a shout for a throw-in, a hard tackle, someone's bag hitting the camera — and deleting one is a single click. Discard as soon as you can tell a clip is not what you came for.

Step 4 — Trim each keeper

Ten seconds is a starting point, not a verdict. Put a clip into edit mode with its Edit control, then drag the in and out handles on the timeline until the clip starts where the move starts. A goal from a corner often wants a longer run-up; a penalty usually wants a shorter one.

One thing the handles do not tell you: when a clip exports by copying (the fast route, and the usual one for an MP4 with no burned-in overlay), the cut has to begin at the keyframe at or before your in point, so the file can open a second or two earlier than where you dragged. The end lands where you asked. There is more on this in best format and settings for football match footage.

If your recording has a noticeable lag between the action and the crowd reaction — a big pitch, a distant camera — you will find yourself pulling the in point earlier on most clips. That is normal and it is consistent within one recording, so after three or four clips you will know the offset by feel.

Step 5 — Export

Pick the tags you want and export. Cutting happens in your browser, so the speed depends on your machine rather than on a queue somewhere.

Three practical notes:

• Exporting the automatically found clips requires an account. Adding a match and previewing the suggestions does not.
• Asking for more than 20 clips in one go raises a confirmation first. On a long match it is easy to have 40 clips still selected without meaning to.
• A batch export writes one file per clip and works through them in order, so a large source file does not normally have to be held in memory — each clip is read straight out of the original. Where the browser cannot do that, the export falls back to loading the whole source into memory and a multi-GB file may fail; exporting a few clips at a time gets around it.

What this approach will not do

It does not understand football. Nothing here recognises a striker, an offside line or a tactical pattern — it finds the moments people reacted to, and then you decide what they were. On a recording with no crowd, no coach and no whistle, there is very little for it to find, and you are better off marking clips manually.

That honesty cuts both ways. On a recording where the touchline reacts, the moments you are looking for are usually the loud ones, so the candidate list is a far better place to start than a raw timeline — how much of the match it catches depends on your own footage, and reviewing the list is how you find out.

Clipping, trimming and exporting are free to use today — see pricing for what the paid packs cover. When you are ready, open AutoHighlight and try it on your last match.