"AI cinema" is used to describe everything from a fully generated short to a conventional film that used a machine-learning denoiser in post. That range makes the term almost useless in conversation. A more practical definition: AI cinema is work where generative or machine-learning tools materially shape what ends up on screen — not just how efficiently it got there.

What actually changes in production

The honest answer is that it changes some stages a great deal and others barely at all.

  • Development and pre-visualisation change most. Concepts that once needed a storyboard artist and a week now take an afternoon, so more ideas get tested before money is committed.
  • Production design and world-building change substantially. Imagery that implied a large VFX budget is now reachable by a small team.
  • Post-production changes unevenly. Cleanup, rotoscoping, and upscaling are transformed; editorial judgement is not.
  • Story, performance, and direction change least. Deciding what is worth showing remains entirely human work, and it is still the part that separates good films from bad ones.

The questions nobody has settled

Authorship is unresolved. When a model trained on existing work produces a frame, the question of what was contributed and by whom does not have a clean answer, and different jurisdictions are arriving at different ones.

Labour is unresolved. These tools remove specific tasks that specific people were paid to do. Any honest account of this shift has to sit with that rather than route around it.

Taste is the quiet problem. When production cost collapses, volume explodes and the scarce resource becomes judgement — knowing what deserves to exist. This is why curation and festivals matter more in this category, not less.

How to actually start

Pick one bottleneck in work you already do and attack only that. Filmmakers who try to replace their whole pipeline at once produce worse films slower. Those who use these tools to solve one real problem — previz, a difficult cleanup, an impossible establishing shot — tend to keep the gains.

Then show the work to people who will tell you the truth about it. Which is, more or less, why VIA exists.

Where to see this work in 2026

VIA runs a three-city Italian season: the Rome AI Festival on 19 September, the Turin Hackathon on 10 October, and Florence AI Cinema Night on 15 November. Between them they cover watching the work, building it, and putting it in front of an audience.