The tools are extraordinary. The output is mostly forgettable, and the reason has nothing to do with the technology.
I use these tools daily and have shipped work that depends on them. The technology stopped being the bottleneck a while ago, which is why the output being mostly forgettable is worth explaining properly.
Cost was doing editorial work nobody credited
When an image was expensive, it had to survive a budget, a schedule and forty people before it existed. That process killed a great many weak ideas without anyone framing it as quality control. It was quality control.
Generation collapses that cost to near zero. The weak ideas now get made, at volume, and arrive looking finished. Finished is not the same as good, but it is much harder to argue with in a meeting.
The tell is not technical
People assume you spot AI work by artefacts — hands, text, physics. Increasingly you do not. You spot it because it has no point of view. Every shot is the most probable shot. The composition is the average of ten thousand compositions of that subject, which is exactly what the model was asked to produce.
That averageness is the actual problem, and no amount of model improvement fixes it, because it is not an error. It is the objective.
Where it genuinely earns its place
Previsualisation, first and most obviously. Showing a client the treatment rather than describing it changes the approval conversation completely, and it has saved me entire shoot days of misunderstanding.
Set extension, cleanup, and the unglamorous middle of post, where it is straightforwardly better and cheaper than the alternative. Variants for testing. And ideation, provided you treat the output as a starting position to argue with rather than an answer.
Where it does not
Anywhere the value of the work is that a specific person made a specific decision. Which, in commercial film, is most of the parts that matter.
The line I keep coming back to
AI is a very fast assistant. Taste is still our problem — and taste is precisely the part that does not scale, cannot be prompted, and is the only remaining reason to hire one director over another.
The studios that will struggle are not the ones who refused to adopt it. They are the ones who adopted it and had nothing else to offer.
- Opinion