AI Filmmaking7 min read6 September 2026
What AI filmmaking actually changed for us — and what it didn't
Two years into building an AI-first pipeline, here is an honest account of where the tools genuinely earned their place, and where they still cost more than they save.
MMohsen
When we started folding generative tools into the studio, the pitch we kept hearing was that everything would get faster. That turned out to be half true, and the half that isn't true is the more interesting half.
What genuinely changed
Previsualisation stopped being a luxury. A director's idea used to live in a paragraph and a mood board until the budget said whether we could build it. Now a rough version exists in an afternoon. Decisions that used to happen after money was committed now happen before — and that is worth more than the render time we saved.
Iteration count went up, not time per shot. We do not finish shots faster. We finish better shots, because we can afford to throw away five versions.
Small clients became viable. A masjid with a modest budget and a good script can now get something that looks like it cost ten times what it did.
What did not change
Everything upstream of the tools. A weak script produces a beautifully lit, expensively graded, weak film.
The model has no opinion about whether your story is worth telling. That part is still entirely yours.
Sound is still sound. Colour is still colour. Continuity is still a person watching every frame with a list. And review — someone who knows the subject checking that what we made is right, not just beautiful — takes exactly as long as it always did. For the work we do, that review is the step we protect hardest.
How we actually work now
- Script and reference first. Nothing goes near a model until the script is signed off.
- Previs everything. Cheap, fast, disposable.
- Generate in passes, not in finals — plates, characters and effects stay separable.
- Grade and sound in a conventional finishing suite. This has not changed and we do not expect it to.
- Review with someone who knows the subject matter. Always.
If you take one thing from this: treat generative tools as a very fast, very literal-minded crew member. Brief them properly and they are extraordinary. Leave the brief vague and they will confidently hand you something wrong.