An automatic highlight video maker can mean two different machines, and picking the wrong one wastes a season. One finds the moments inside raw game footage. The other takes moments you have already chosen and turns each into a finished, branded video from your data. We build the second kind, and this article is about how it works, what it needs from you, and where it stops.
If you run digital for a club, a college athletics department, a league, or a streaming outfit, the second machine is probably the one that solves your actual week. The footage is cut. The stats exist in a sheet or a feed. What you lack is hands to produce forty player videos before the next game starts. Here is how that job gets done without forty trips through the same After Effects project.
Two meanings of “automatic highlight maker”
Search for a highlight maker and you will find tools that promise to watch a full game and cut the best plays on their own. That is a computer vision problem. Software has to decide what a good play looks like, find it in hours of video, and trim around it. Some of that work is real and improving. We do not make a tool that does it, and it is worth being plain about that so nobody buys Templater expecting it.
The other meaning is less glamorous and, for most production teams, more useful. Someone, or something, has already picked the clips. A scout flagged them, a replay operator logged them, a clip library holds them with names that match the players. The remaining work is packaging. Put the player’s name over the clip, add the stat line from tonight’s game, drop in the team colors, match the sponsor tag, render a vertical version for social and a wide one for the stadium screen. Every one of those steps is the same for every player. Only the data changes.
That packaging work is what a data-driven highlight maker automates. You design the look once in After Effects. You link the parts that change to a spreadsheet or a data feed. Templater then renders one finished video for each row. A row about one player becomes one video about that player.
Two Different Tools
Clip-finder vs. clip-packager: pick the right one
- What it starts with
- Clip-detection toolHours of raw game footage
- Data-driven packagerClips already chosen and named
- Core task
- Clip-detection toolFinds and trims the best plays
- Data-driven packagerPackages clips into branded videos
- What drives it
- Clip-detection toolComputer vision on video
- Data-driven packagerA spreadsheet or data feed
If you already have the clips and the stats and your bottleneck is turning them into dozens of branded videos on a deadline, a data-driven approach fits.
The two approaches also combine well, and this is where the interesting work is heading. In a post about AI in sports video, we pointed to University of Waterloo research on tracking hockey games with AI. The researchers reported that their system identified teams with 97 percent accuracy and individual players with 83 percent accuracy. A system like that produces data. Data still has to become something a fan can watch. We described Templater as the glue between the two: the AI output goes in as rows, and the videos come out. If you want to know where we draw the line on AI more generally, our post on what Templater AI does and does not do spells it out.
One game night, built from a single sheet
Picture a university athletics department with a basketball program that plays at home on Tuesday night. The social producer wants a short highlight card for each player who scored, posted within an hour of the final buzz. Fifteen players might qualify on a good night. Doing that by hand means opening the project, retyping a name and stat line, swapping the clip, adjusting text that no longer fits, exporting, and starting over. By the fifth player the producer is tired and the typos begin.
Here is the same night built as a data job.
The template. A motion designer builds one composition. It has a text layer for the player’s name, one or two for the stat line, a placeholder for the headshot, a placeholder for the highlight clip, and a lower section for the sponsor. The designer builds it once, in the school’s colors, and it looks the way the brand guide says it should. After that, nobody touches the design on game night.
The sheet. Each row is a player. Columns hold the name, the stat line, the path to the headshot, the path to the clip, and an output name. Templater reads Google Sheets, tab-separated files exported from a spreadsheet app, JSON files, and JSON feeds. A sports information director who has never opened After Effects can fill in a Google Sheet. That is the point. The person closest to the stats owns the data. The designer owns the template. Neither needs the other at ten o’clock at night.
If you want a deeper look at how to structure that sheet, our guide to setting up a roster sheet that builds every game’s player graphics walks through it. And for the speed side of the story, posting score and highlight graphics minutes after the final whistle covers the timing.
The run. Templater works through the rows. Each row is applied to the master composition, and the result is a finished video. One row, one video. If the producer wants a second format, a square cut for one platform and a vertical one for another, those become additional rows or additional outputs in the same run. You do not rebuild anything.
Game Night Workflow
Three steps turn a sheet into player videos
-
The templateA motion designer builds one composition, text layer for the player’s name, stat line, headshot placeholder, clip placeholder, and sponsor section, in the school’s colors.
-
The sheetEach row is a player.
-
The runTemplater works through the rows.
The clips are the part people worry about first, because every clip has a different length. A three-second dunk and a nine-second possession cannot sit in the same fixed layout without something giving. Templater has a feature called Time Sculpting for this. It crops, trims, shifts, and stretches layers so footage of varying duration fits the design, and Dataclay holds a patent on it. You set the rules once in the template. Each clip then finds its own fit. We cannot promise it will make a weak clip look good. It can make sure a short clip and a long clip both land inside the same graphic frame without an editor trimming each by hand.
What goes wrong, and the habits that catch it
Automation does not remove mistakes. It repeats them quickly. A wrong font size on one video is a five-minute fix. The same wrong size across fifteen players, or across a whole conference’s worth of rows, is a long render you throw away. The good news is that the risks in a highlight job are predictable, so you can design for them.
Long names and short names. The surname that is twice as long as any other will break a layout built around “Lee.” General practice is to test the template against the longest and shortest entries in your data before game night, not after. Put the awkward names at the top of a test sheet on purpose.
Missing media. A headshot that was never uploaded, or a clip file with a typo in its path, is the most common reason a batch stalls. Templater has a setting that decides what happens when a footage reference in your data points at nothing, and the right choice for a large run is to let a branded placeholder stand in so the batch keeps moving. You then fix the two or three rows that used the placeholder and run only those. We cover the setting in handling missing footage and remote footage in Templater.
Data that is right but wrong. A stat that was corrected at halftime. A name spelled one way in the roster and another on the broadcast. The sheet will hand both to the template, and the template will render them faithfully. Faithful is not the same as correct. This is why a person has to look before anything goes public.
That person has tools. Templater lets you preview a composition with your data before you render, stepping through the rows to see how the template treats different text, footage, and colors. For bigger runs there are spot checks, which export still images or short clips from chosen frames of the data-merged compositions. You mark frames with named timeline markers, or Templater samples frames evenly across the duration if there are none. Spot checking needs Templater 3.0 or newer and works in the Pro and Bot editions. Our post on preview modes and spot checks covers the order of operations.
Here is a point we want to be fair about. Research we have written about on automated news video found that highly automated videos were liked significantly less than human-made ones, while partly automated videos that people edited afterward were not liked significantly differently. That study looked at news video, not sports. We would not claim it proves anything about a dunk reel. It does match what we see as a sensible practice. Let the template and the data do the repetitive work, and give a human one look at the output before it posts. For sports, we think that look is cheap insurance, because the cost of putting the wrong player’s name on a highlight is far higher than the time it takes to glance at a preview.
How the batch starts, and how it gets out the door
Templater comes in three editions. Rig is for building and rigging templates. Pro renders or replicates a batch when someone clicks the button. Bot watches the data source and starts customizing and rendering on its own when new data appears, and it can also take data pushed in from the command line. For highlights, the choice is a question of how much you want a person in the loop.
Edition Comparison
Three editions match how much oversight you want
| Edition | What it does | How a batch starts |
|---|---|---|
| Rig | Building and rigging templates | Manual, design phase only |
| Pro | Renders or replicates a batch | Someone clicks the button |
| Bot | Customizes and renders on its own | New data appears, or data pushed from command line |
A department that wants a producer to approve every night’s package can run Pro. The click is a built-in checkpoint. A team that posts routine recaps all season and trusts its sheet can run Bot, with spot checks reviewed on a schedule. Same template, different supervision. You can pick by how much a mistake would cost on that particular piece of content. A season-opening hype video deserves a human. A weekly leaderboard card probably does not need one.
There is a third path worth knowing about. Instead of rendering at once, Templater can replicate each row into its own composition inside After Effects. A designer can then adjust the few that need hand finishing, and send everything to the After Effects render queue or to Adobe Media Encoder. This is how you handle the one video in fifteen that needs a custom touch without giving up the other fourteen.
Rendering is where batches slow down. After Effects renders one video at a time. Projector, our rendering and distribution app, renders several customized videos at the same time, based on the number of cores in the machine. Dataclay’s own example is that a quad-core machine can process four videos in roughly the time After Effects takes for one. It can also transcode as part of each render and deliver files to Vimeo, JW Player, Frame.io, and AWS S3. It can produce GIF clips and poster images, which the social team usually wants anyway, and it writes metadata from the streaming platform back to your data source. It runs on your own machine, so unreleased footage stays under your control.
We should be careful about speed claims. How long your batch takes depends on your machine, your template, and your footage, and we have no benchmark to quote for a highlight job. The sensible way to learn is to run your own template on your own hardware. The Templater trial lasts 30 days and gives you the full Bot edition, so you can test the whole chain, from sheet to rendered file, before deciding anything.
Where this is the wrong tool
Some honest limits, because we would rather you hear them from us.
If your problem is finding the moments, this is not your tool. Templater does not watch a game and pick plays. If you have hours of raw footage and no one logging the good parts, you need a clip-detection product or a person, and you should evaluate those on their own terms. Once the clips exist and are named, we pick up from there.
If nobody on your team works in After Effects, someone has to build the first template. That is a real cost. Dataclay also offers custom development for teams in that position, and Autografs are pre-built After Effects compositions already rigged with Templater, which can shorten the start.
If each highlight is a one-off with unique editing, such as a feature piece with narration and hand-timed cuts, a template does little for you. The payoff comes from repetition. The more videos share one design, the more a template earns its keep. A rough rule of general practice: if you would make the same video more than a few times with different names and numbers, it is a candidate. If every video is its own creative problem, it is not.
And the software will not decide what deserves airtime. A sheet can say who scored. It cannot say whether the moment calls for a loud treatment or a quiet one. Our post on which parts of a sports broadcast workflow can be templated draws that line in more detail.
The bottom line
Decide which highlight maker you need before you shop. If you need software that finds the moments in raw footage, that is a different product from ours. If you already have the clips and the stats and your bottleneck is turning them into dozens of branded videos on a deadline, a data-driven approach fits. Build one template. Keep the stats in a sheet or a feed. Preview before you commit, spot check the big runs, and let a person look at the output before it goes live. The design stays fixed and the data does the changing, so a correction on game night is an edit to a cell, not a trip back into the project.
Start with a single game and a single template. If the first night works, the second is mostly a new sheet.
We covered the main points above. Here are the questions producers ask us most when they are weighing this approach.
Frequently Asked Questions
Does Templater find the best plays in a full game video?
No. Templater does not watch footage and pick moments. It packages clips you have already chosen. It places each clip into an After Effects template, adds data such as a player name and stat line from your sheet or feed, and renders one finished video per row. If you need automatic play detection, that is a separate kind of tool, and Templater can take its output as data.
What data sources can drive a highlight video?
Templater reads Google Sheets, tab-separated files exported from a spreadsheet app, JSON files, and JSON feeds. For larger setups, Que stores each job as a JSON object in the cloud and accepts data through an API or through Zapier, which connects to more than 5,000 apps. A person typing into a sheet, a form, or another system can all supply the rows.
How do clips of different lengths fit the same template?
Templater includes a feature called Time Sculpting, which crops, trims, shifts, and stretches layers so footage of varying duration fits the design. You set the rules once in the template, and each clip adapts to them. It handles fit, not quality, so a weak clip will still look like a weak clip once it is trimmed to size.
What happens if a headshot or clip file is missing?
Templater has a setting that controls what happens when a footage reference in your data points at nothing. For a large batch, the best choice is usually to let a branded placeholder stand in, so the run finishes. You then fix the few rows that used the placeholder and run only those, instead of discovering the problem hours later.
Do I need a person to review each video before it posts?
We recommend a look before anything goes public, though how closely depends on the stakes. Templater offers composition previews and spot checks for this. The Pro edition starts a batch when someone clicks, which is a natural checkpoint. Bot starts renders on its own, so pair it with spot checks that someone actually opens and reads.
Can renders start automatically when new stats arrive?
Yes, with the Bot edition. Bot checks the data source for new data and starts customizing and rendering by itself, and it can also take data pushed in from the command line. The Pro edition waits for someone to click. Many teams use both, matching the edition to how costly a mistake would be on each type of video.
Can I try it before committing?
Yes. The 30-day Templater trial is the full Bot edition, so you can run a real template against a real sheet and watch the whole chain work. Use your own longest player name, your messiest clip, and a missing headshot on purpose. Whatever breaks in the trial is cheaper to learn than whatever breaks on game night.
