Taste is not the edit
Watch ten Christopher Nolan films and you feel something before you can name it. Something Nolan-y. It is not his cuts. It is how he sees time, and guilt, and obsession. The cuts are just where you catch it.
That feeling has a name. We call it taste. But we use the word wrong. We say a film has taste, an edit has taste, a painting has taste. None of that is true. The edit is not the taste. The edit is the delivery. Taste is the thing underneath, the thing it all leaks out of.
The clearest proof of that is a man who has been dead for five hundred years.
Da Vinci was not painting pretty
His paintings are not famous because they are pretty. They are famous because they show you reality. He was obsessed with anatomy, with light, with physics, and most of all with the eye. He called the eye the window of the soul. He was the first person to work out that the eye works like a camera, that the image lands upside down inside it. Those soft, melting edges in the Mona Lisa are not a style. That is just how the eye actually sees an edge. He even figured out that the pupil opens up at dusk, so he said that is the best time to paint.
So the painting was never really the point. The obsession with how we see was the point. The painting is just what fell out of it.
That is taste. It is how you view the world. And you cannot really change it. You can change your tools, your software, the trend you are chasing this week. You cannot change how you see. Which left us with a strange question.
Is there a pattern to it?
If taste is a worldview, and a worldview shows up the same way again and again, then is there a pattern to it? Some shape, some math of taste, the way people believe there is a math of creativity? We were not sure. But it looked like it repeats. And if it repeats, you can study it.
One thing first. There is no good taste or bad taste. We are not chasing good. We are chasing the taste a lot of people quietly agree on. The socially accepted one. Because that is the one you can repeat. And to repeat something, you have to see it clearly first. For that, you need data.
Turns out people have been trying to crack this for a hundred years.
People tried to write the formula
In 1933 a mathematician named Birkhoff wrote down an actual formula for beauty. Beauty equals order divided by complexity. Clean. Simple. And it did not work. Every time someone tested it on real people, it fell apart. Taste is not an equation you can write on a napkin.
But the thing that replaced the formula is the interesting part. The most attractive face, it turns out, is just the average face. Take a thousand faces, average them together, and people find that one more beautiful than almost any single face in the pile. Beauty sits at the center of the data. Caricature is the opposite trick. You take the way a face is different from the average and you push it further, and somehow it looks more like the person than the person does. So art is the average, plus the right amount of difference from it. You cannot guess those numbers. You only find them by looking at a lot of examples.
Even the old philosophers landed in the same place. Hume said taste feels personal, but when enough good eyes look at the same thing over enough time, they start to agree, and that agreement becomes the standard. Bourdieu said taste is social, something a group builds together. Every one of them is describing the same thing. A standard nobody can write down, but that you can find inside a crowd. And a crowd is just a dataset.
So we do not write taste in. We grow it.
That is the bet behind Kinetic. You cannot write taste into a machine as a rule. We tried. Everyone tried. It does not hold. So we stopped trying to write it, and we started growing it from data. Three real parts do the work.
The first is a library. A folder of references, and each one is saved so the engine can actually build it, not just admire it. A reference is not a video we copy. It is a recipe. The colors, the pace, the kind of caption, the type of motion, all stored as something the renderer understands.
The second is a lane. A small file that holds the taste of one person or one brand. Fonts, colors, how fast things move, how loud the sound sits, what counts as too much. Change the file, change the edit. Same footage, different eyes.
The third is a loop. Every time you keep an edit, kill it, or change it, that choice gets logged. Over enough of them, the system starts to guess what you would have done before you tell it. That log is the thing nobody else has. That is the moat.
The bet
A machine can copy a cut in a second. It will never copy how you see the world. But if we can capture enough of how you see, and hold it in data, then it can make the cut you would have made.
None of this is finished. We are early. But the shape is right, and the shape is the bet. That is the whole company.
We build AI that has a point of view
Kinetic is ours. If you have an AI product that needs to actually work, not just demo well, come talk to us.
Get in touch