Process
How we make 100,000 personalized videos
One film, a hundred thousand versions, each one addressed to a real person by name. Here is how it is built.
A few years ago, sending 100,000 people their own video would have meant 100,000 shoots, which is to say it would not have happened. Today it is a single idea, made once with real care, then turned into a hundred thousand individually addressed versions by a pipeline that runs while we sleep. We built one of these for a global crop-science brand. This is how the machine works, without the client's name on it.
The short version: you do not make 100,000 videos. You make one excellent film and one excellent system, and the system makes the rest.
Why one to one beats one to many
A broadcast video is built for an average person who does not exist. It is polished, it is general, and it asks everyone to find themselves somewhere inside it. Most people do not bother. A one-to-one video does the opposite. It speaks to the actual viewer, by name, about the thing they actually care about, in the language they actually think in. Relevance is the price of attention, and a personalised film pays it up front. That is the whole bet, and the numbers reward it. It is AI-driven personalization at its most literal: the creative itself made one to one, not just the targeting around it.
The pipeline, step by step
It starts with a master film. This is the only part that is made the old, slow, expensive way, and that is deliberate. A director, a real edit, considered sound, proper craft. Everything downstream inherits its quality, so the master has to be genuinely good before a single version is rendered. If the master is mediocre, you have simply industrialised mediocrity.
Next comes the data. Clean, structured data is what decides who gets what: a name, a region, a crop, a stage in the season, a preferred language. Personalisation is only as honest as the data behind it, so we spend real time mapping fields and removing the gaps that would otherwise produce a video addressed to "Dear Valued Customer".
Then the templating and AI production layer. This is where the master film and the data meet. The film is broken into fixed parts that never change and variable parts that do, and the variable parts are generated for each record. A name spoken aloud, a line of on-screen text, a localised shot. The system assembles each version frame by frame, file by file - personalized video production at scale, run as one pipeline.
After that, language and voice variants. A campaign that runs across ten languages cannot simply subtitle one English film and call it localisation. We generate voice and text in each language so the video sounds native, not translated. We have written more about holding craft across many languages in craft at scale.
You do not make a hundred thousand videos. You make one beautiful film and one disciplined system, and the system protects the film at scale.
Then quality control, which at this volume cannot be a human watching every clip. Automated checks confirm that each file rendered fully, that the right name landed in the right place, that the audio and video are in sync and that nothing is broken or blank. The rare failures are flagged for a person to look at. The rest pass clean.
Finally, delivery. A finished video is worthless sitting on a server. Each version goes to the right person on the right channel, whether that is a messaging app, an email or a link, timed to the moment it is most useful.
How craft survives the scale
The fear with anything at this volume is that it will feel like volume. It does not, and the reason is structural. The craft lives in the master, which is made by people to a high standard. The system does not invent quality; it preserves and repeats it. Once the master is right, the machine simply refuses to let the standard drop on version 90,000. Scale is not the enemy of craft here. The system is the thing that defends the craft.
What makes it pay back
Three numbers move when this is done well. Relevance goes up, because each film is built for one person. Completion goes up, because people watch something made for them to the end. And response goes up, because a relevant message at the right moment is far harder to ignore than a general one. That is the difference between a film that wins attention and a film that grows the business. Personalisation at scale is not a gimmick. It is the most efficient way we know to make one good idea matter to a hundred thousand different people.
Questions, answered
How do you make 100,000 personalized videos?
You start with one strong master film, then build a system around it. Clean data supplies the variables, a templating and AI production layer renders the changing parts, language and voice variants cover each audience, automated quality control checks every file, and each finished video is delivered to the right person on the right channel.
What does it take to personalise video at scale?
Three things working together: a beautifully made master film that holds the craft, clean and well-structured data that decides what changes for whom, and a production pipeline that can render, check and deliver tens of thousands of versions without a person touching each one.
Does personalized video actually work?
Yes, when the personalisation is genuinely relevant rather than a name dropped into a generic clip. Relevance earns attention, so completion and response tend to rise sharply against one-to-many video. The test is simple: does each version feel like it was made for the person watching it.