Where AI Films Die.

You've seen the bad AI films. So have we. We've made a few. We didn't release them, which is the entire difference, and it's a bigger difference than it sounds.

Here is the uncomfortable part: most of those films didn't fail where everyone thinks they failed. The generation was fine. The faces held. The physics behaved. They died afterwards, in the part of the process nobody tweets about — the part that doesn't demo well.

They died in post.

A generated shot arrives like an ungraded dailies roll: technically an image, emotionally nothing yet. On a traditional job nobody would dream of airing dailies. In this category, people air dailies every day, with a music-library track underneath, and then wonder why the film feels like a screensaver with ambitions.

Most bad work in this category fails in post.

The craft that fixes this is not new, and that's the point. A colourist who pushes the highlights until the frame has weather. A composer who scores the cut instead of decorating it. An editor who kills a beautiful shot because it's four frames too long and beauty was never the assignment. None of these people care how the pixels were born. They care whether the film breathes.

Anyone can generate an image. There are teenagers doing it on a phone, and some of them are better at it than agencies charging six figures. Knowing which image deserves to exist — and killing the thirty-nine that don't — is a different profession. It's the one we've been practising for twenty-five years, mostly on films that were shot on actual cameras with actual overtime.

So when people ask what tools we use, we tell them, and it never satisfies anyone, because the honest answer is: the tools you'd find in any serious cutting room, plus some new ones in front. The last mile of every film we make is a grade, a mix and an argument about taste.

The argument is the product. Everything else is rendering.