Designing a cricket video editor around the bat-ball moment

Most "cricket highlights" get made the hard way: you film a net session or a match, then sit at a phone or a laptop for an evening, dragging a scrubber back and forth to find the good balls. The tool doesn't know anything about cricket — to it, your footage is just pixels and a waveform. Every decision is yours, one frame at a time.

CricCuts started from the opposite assumption. If you're building a cricket highlights app — a true cricket highlights creator, not a generic trimmer — you can bake the sport into the product. You know there'll be deliveries. You know each one has a moment of contact. You know the interesting thing is the shot, not the walk back to the mark. Once the app understands that shape, almost every UX decision changes. This post walks through the ones that mattered most.

Why a cricket video editor can't just be a generic trimmer

Open a normal cricket video trimmer and it treats a two-hour recording as one undifferentiated block. You get a timeline, a playhead and your own patience. Finding thirty good balls in a session means watching most of the session. The editing isn't the hard part — the finding is.

A cricket-specific editor can do the finding for you. CricCuts watches the footage, picks out the deliveries that look and sound like real shots, and lays them out as clips you can review in a couple of minutes. It's how to make cricket highlights without an evening of scrubbing: let the app auto cut cricket highlights for you, then spend your time choosing shots rather than hunting for them. But automation is only half the job. If the app is going to hand you a stack of clips, every screen has to make one thing effortless: knowing, at a glance, which ball each clip is. That single requirement drove the design of almost everything you see.

The goal was never "one clever algorithm." It was a product where a cricketer opens the app after a session and, without reading a manual, gets a shareable reel of their best shots — and trusts it.

Every thumbnail is framed on the bat-ball moment

Here's a small decision that quietly runs through the whole app. In most editors, a clip's thumbnail is its first frame — which, for a cricket shot, is usually the bowler at the top of their mark or the batter still walking in. A wall of those thumbnails all look identical. You can't tell your cover drive from your pull without playing each one.

So in CricCuts, a shot's thumbnail is pinned to the moment of contact — bat meeting ball — not the first frame and not a random one. The still you see is the shot. Scroll the timeline, the clip list, the curation cards or the review-and-trim screen and every tile reads instantly, because each one is frozen on the actual stroke rather than the run-up before it or the celebration after. A duration badge (say, 3s) sits in the corner so you know the clip's length without opening it.

There's one deliberate exception. When you're choosing which batter to follow, the app parks on a clean "standing ready" frame a beat before the swing — because a frame mid-stroke is motion-blurred and hard to draw a box around. You can nudge frame-by-frame to land on exactly the still you want. And when you actually play a clip, it opens a moment before the action, so you watch the whole shot develop. Only the thumbnail is pinned to the peak; playback always gives you the lead-in.

Reading the ground: adjacent nets, any surface, and a batter who won't stand still

The hardest part of a real net practice video editor isn't a clean, staged clip — it's the messy reality you actually film in. Three problems show up in almost every session, and each one needed a product answer, not just an engineering one.

Action in the next net over

Film in a multi-net facility and there's a game happening on either side of yours. A naïve editor cheerfully clips the bloke two nets down every time he middles one. CricCuts asks you a single, plain-English question up front — is it just your batter, or are others playing nearby? — and when you pick "others playing," it works to keep your batter's shots and drop the neighbouring net's. You mark your player once, and the app stays focused on your lane. (The story of how it learned to tell nets apart is its own post: the idea the AI missed.)

Any surface, any angle — without a settings menu

Cricket gets played on everything: turf, matting, astro, concrete, indoors under lights, outdoors in the wind. Rather than make you classify your pitch, the analysis calibrates to each recording you bring in, so it reads cleanly across astro, matting and turf with no setup. The one thing worth telling it is where your phone was — behind the bowler, side-on, or behind the batter — because that genuinely changes what a shot looks like, and it's a two-second tap.

The batter who taps and wanders

Watch any batter get ready and they tap the bat on the ground, shuffle their feet, walk out to the pitch of the ball, come back. A tool that fired a new clip on every tap and every footstep would bury you in fragments. CricCuts groups all the little sounds of a single delivery into one clip per shot, anchored on the actual contact rather than a preceding tap or the bowler's approach — and it keeps following your batter as they move around the crease, so it stays locked on the right person. A gentle bat-tap with no real shot behind it gets left out of the recommended picks instead of cluttering your reel.

It started as a data-science tool — then we began taking things away

There are two ways to deal with messy footage like that. You can hand the user a pile of filters to clean it up afterwards, or you can design the tool so the mess rarely reaches them in the first place. CricCuts started with the first approach — and has spent its life moving to the second.

The honest origin story is that early CricCuts was closer to a data-science instrument than a cricket app. When your engine produces a lot of signals, it's tempting to put every one of them on screen. So there was a dedicated detection "fine-tune" screen, and a filter panel that was essentially a wall of sliders — each governing a different technical signal, each labelled in the language of signal processing. Powerful, if you enjoy tuning a detector. Baffling if you just wanted your cover drives back.

The lesson took a while to land: a cricketer doesn't want to operate a shot detector — they want their shots. Every extra knob is a decision you offload onto someone who opened the app precisely to avoid decisions. Advanced filters don't make a tool smarter; they make being smart the user's job.

And the subtraction isn't finished. The last cluster of power-user controls — an "advanced filters" panel of score cut-offs and hide-below-this sliders — is the next thing on the chopping block. If a sensible default does the job for the vast majority of sessions, the slider shouldn't exist at all. The intelligence belongs in the product, not in a settings screen: a simplified tool isn't a dumber one, it's one that has already made the hard calls so a cricketer gets there in four taps, not forty. It's the difference between a cricket video editor built for the player and one built for the engineer who made it.

The boundary box: the one thing we ask you to do — and why it keeps the data honest

We took away almost every knob. We deliberately kept — and will defend — exactly one small ask. Before CricCuts studies the video, it lets you point out your batter: scrub to a frame where they're standing ready, and draw a box around them.

Why keep this, when we removed everything else? Because it isn't a filter — it's the opposite of one. A filter cleans up bad results after the fact; a box prevents them. Any tool built on data lives or dies by what goes into it — garbage in, garbage out — and one accurate mark up front tells the analysis exactly which player to follow. Every clip it then produces is tied to the right batter: it locks onto your net and leaves the shots flying in the net next door alone.

It's data accuracy by construction: fix the input once instead of filtering the output forever. That single box is why we could delete a dozen clean-up filters and end up with a more accurate tool, not a less accurate one.

How a box enforces accuracy

The clever part is the check. CricCuts won't let you move on until it confirms a real batter is actually inside your box — a green tick means it's locked on; a red cross tells you to redraw it tighter and taller, or scrub to a cleaner frame. And that confirmation looks at the frame the same way the full analysis will, so a box that passes here is one the analysis can genuinely work with. You never kick off a long, phone-side analysis pointed at an empty or wrong box — the quality gate is baked into the step.

Even the on-screen guidance is a data-quality lesson in disguise: it nudges you to draw a tall, snug box around just your batter, and warns that too wide a box "pulls in the adjacent net." That's the whole philosophy in one sentence — give the analysis a clean, unambiguous input, and everything downstream stays clean.

Two things worth being clear about. It's where, not who: there's no face recognition and nothing is uploaded — you're simply telling the app which player to follow, and that one mark is reused across the whole session. And it's optional: don't want to draw it? Auto-detect finds the batter for you to confirm, or Skip it entirely and your highlights still come from the sound of the game. Even the one ask we kept, we made skippable — the tool should do the work, and only ask when your input makes the data meaningfully better.

Two ways to cut: TokTok bursts vs. keeping the coaching in

A cricketer making a reel for Instagram wants something very different from a coach reviewing a session with a player. Same footage, opposite intent — so the app lets you pick how each clip is cut, before it even starts.

These aren't buried in an export dialog — they're the first thing you choose, because they change what the whole reel is. TokTok is the cricket reel maker for showing off; longer clips with the audio left in make it a cricket coaching video editor for getting better. A clip-length control (short and snappy through to several seconds) and a target reel length round it out — set a total, and it packs your best moments to fit. And because the one-second TokTok cut is also a one-tap option on the review screen, you can punch up a longer edit into a reel without re-analysing anything.

Why does watching yourself back matter so much for a coaching cut? That's a rabbit hole worth falling into — the neuroscience of learning from watching yourself is the reason the "keep the coaching in" option exists at all.

A loading screen that makes you a better cricketer

Analysing a full session takes a little time — the work is real, and it all happens on your phone, not in a data centre. A spinner would have been the easy call. Instead, the wait became one of the most-loved parts of the app.

While CricCuts works, it shows a rotating "scorecard" of cricket content you can flip between: training tips, did-you-know facts, the laws of the game, and short life & leadership lessons from cricket history — over 400 cards in all, shuffled so you rarely see a repeat. Cards drift past on their own, or you can swipe through by hand. It's genuinely useful downtime: you might learn that the mock obituary printed after England's 1882 loss is where "the Ashes" comes from, that a "duck" is short for duck's-egg, or a coaching cue like "watch the ball out of the hand — track it from the bowler's grip all the way to the bat."

🏏 A friendly line tells you where the analysis is — "Listening to your match…", "Finding the shots, appeals and big moments…", "Finishing your reel…" — over a cricket-bat-shaped progress bar that fills from handle to toe with a little red ball riding the leading edge. The wait tells you what it's doing, and leaves you a bit sharper.

It's a small thing that says something bigger about the product: the time your phone spends thinking shouldn't be dead time. (It runs entirely on-device, which is a deliberate choice in itself — here's why the future of AI is small models on the edge.)

Getting noticed — and staying yours

A highlight only matters if it gets seen by the right people: a coach, a selector, a club, your feed. So the export step is built to do two jobs at once — help your clips travel, and make sure they stay attributed to you when they do.

Sized for wherever you're posting

A watermark that carries your name, not a lock

Every exported reel is branded and baked in. The CricCuts mark sits in the corner of every frame; beneath it you can print your Instagram handle, and in the opposite corner a scannable QR code pointing anywhere you like — your Instagram, your CricHeroes profile, your website — so someone who likes a clip can find you in one tap. Because the branding is rendered into the video, it survives downloads and re-posts.

To be straight about it: a watermark can't stop anyone from re-sharing your clip — nothing can, once it's out there. What it does is make sure that when your shot gets passed around the group chat and beyond, your name and your handle travel with it. That's the realistic, useful kind of protection for a young cricketer trying to get noticed: not a lock on the door, but your signature on the work. When it's ready, the reel saves straight to your gallery and shares to any app in a tap.

Designed to be forgiving: you always have the final say

Automation you can't override is stressful, not helpful. The strongest product decision in CricCuts is that nothing is decided for you. Every detected moment lands on a clean, reviewable timeline. Keep or drop any clip with a tap; select or deselect the lot in one go; sort by time or by shot quality. Each clip even wears plain-language quality labels — Power, Appeal, Motion, Run-up — so you understand why it surfaced.

The details are where the care shows. There's real Undo/Redo, so experimenting is risk-free. Your manual picks survive a full re-analysis — change a setting and the clips you personally chose or dropped don't get wiped. Empty screens say something human ("Every detected event is already in your reel.") instead of going blank. A first-run tour walks you through every stage, and a "?" on every screen jumps straight to the right page of the guide. You can start as a guest, and your videos never leave the phone.

Frequently asked questions

Keep reading: the neuroscience of learning from watching yourself, the idea the AI missed, and why the future of AI is small models on the edge. More on the CricCuts blog.