The idea was never the expensive part

There is a box on this site. You type a brand into it — yours, your client's, a competitor's — and about twelve seconds later you have five film concepts, each with a logline, a three‑shot list, a director's look and a production range. It is free. There is no sign‑up. People keep asking what the catch is, which is a fair question to ask a studio that gives away the thing agencies have historically guarded.

There isn't one. The reason is simpler and slightly less flattering to our own industry: the idea was never the expensive part.

What the old arrangement assumed

For most of advertising's history the deck was the asset. Agencies pitched for free, sometimes against five other shops, and the concept was what they were protecting on the way out of the room. That made sense. A concept took a creative team a fortnight, a planner before them and a room of people arguing after. It was genuinely scarce, and scarce things get guarded.

The guarding outlived the scarcity. A speculative concept — a premise, a shape, a line — now takes seconds, and not because anything became clever. A concept is largely recombination: this category, that tension, this device, pointed at a brand. Recombination is the thing these models are unusually good at. Insisting otherwise is a pricing position, not a description of the work.

If a tool can produce it in twelve seconds, it was not the part you were being paid for.

What is actually scarce

Three things, and the tool has none of them.

Knowing which of the five is the one. Our predictor puts a score on each concept. A score is a sorting mechanism, not taste. The judgement that matters is whether an idea is worth a year of a brand's attention and a budget that could have gone to media instead, and nothing about producing the idea gives you that.

Making it. This is the gap people underestimate most, and it has got wider rather than narrower. A logline is one sentence. A film is a face that has to stay the same face across forty shots, a grade that holds from a phone screen to a cinema, sound that was composed rather than selected, and a hundred continuity decisions that no prompt contains. Most of what goes wrong in AI film goes wrong there, not at the idea.

Standing behind it. A concept costs nothing to propose. Someone has to say this one, at this budget, by this date, and be answerable when the client's board asks why. That has never been automatable and is not close to becoming so.

Not all five will be good

We should say this plainly, because you will find out in twelve seconds anyway. Of the five concepts it gives you, one or two will usually be interesting, one will be competent and forgettable, and the rest will be the kind of thing that sounds like advertising because it was assembled from advertising. We would rather you saw that than a curated demo with the misses removed.

The test we use on our own work is blunt and you can apply it to the output in a few seconds. Swap the brand. If the concept still functions with a competitor's name on the end, it is not an idea about that brand, it is a format with a logo attached. The good ones fail that test immediately — they only make sense for the company they were written for, because they are built on something only that company has.

What we chose not to build

The obvious version of this tool asks for your email before it shows you anything. We built that and then took it out.

Partly because it is irritating, and partly because it contradicts the argument. Putting a wall in front of the concepts implies the concepts are the product. They are not. They are the clearest demonstration we can offer of how we think about a brand, which is the actual thing anyone is choosing between studios on. A wall would hide the demonstration in order to protect the sample, and the sample is worth less than the demonstration.

So it runs for anyone, repeatedly, with no form. There is a per‑day ceiling on it because the thing costs us money to run, and that is the whole of the restriction.

The part that is a pitch

This is not altruism and it would be irritating to pretend otherwise. We think that if you use it and one of the five makes you want to see the film, you will have already learnt the only thing a credentials deck is trying to tell you: whether these people think about your brand in a way you find interesting.

And then the conversation starts at the point that actually matters, which is whether it can be made well, for what, by when. We have written elsewhere about what that looks like on our side and what a brief needs to contain before anyone can answer it honestly.

Type a brand into the predictor and see what it says about yours. If all five are wrong, that is useful information about us, free, in twelve seconds.