The most useful thing a studio can offer right now isn't the newest model — it's judgement about where to use one. Our internal rule is simple: AI is excellent at volume, iteration and cleanup, and unreliable at intent, taste and continuity. Everything else follows from knowing which side of that line a task falls on.
What AI is genuinely great at
Generating options fast, exploring look and mood before budget is committed, localising and versioning at scale, and grinding through the unglamorous hours — rotoscoping, rig removal, upscaling, background fills. In these places AI doesn't lower quality; it frees your best people from work that was never creative to begin with.
What stays human
Direction, performance, edit rhythm, sound and final look are where a film is actually won or lost — and they remain firmly human. So does knowing why a choice is right for this brand and this audience. A model can produce a thousand frames; it can't tell you which one is on-brand.
The line: continuity and intent
The clearest failure point is continuity. The moment you need the same product, face or environment to stay identical shot to shot, unsupervised generative tools drift. And intent — the specific thing this piece needs to communicate — is something you bring to the tool, not something you get from it.
Our internal rule, in practice
- Concept and explore with AI — fast, cheap, wide.
- Decide with humans — taste and brand judgement.
- Produce with the right mix — CG, live, or generated per shot.
- Finish with humans — edit, grade, sound, and a final quality bar.
The bottom line
Used well, AI is a way to spend your best people's time on the decisions only they can make. Keep it on the volume-and-cleanup side of the line, keep humans on intent-and-continuity, and you get work that ships faster without looking like a machine made it.