Film and television do more than entertain: they circulate images, gestures and ways of telling stories across borders. In this seventh essay in my 2024 series, I consider how AI could change that process, from writing a scene to deciding which film reaches an audience.

For a writer, a generative system can be a tool for trying alternatives: a scene outline, a change of viewpoint or several versions of a conversation. Analysis of existing scripts can suggest patterns to investigate. But a pattern found in successful films is not a recipe for another success. Dramatic tension, a character’s contradictions and an unexpected silence still require decisions about what the story means.

Casting raises a related question. Data about previous roles or audience response may support research, but cannot establish the chemistry between performers or decide who should embody a character. Digital replicas add another layer. A convincing face or voice is not simply a production asset: its use calls for clear agreement about consent, context and compensation. Technical feasibility does not settle those questions.

In production, AI-assisted tools can help with planning, image preparation and parts of post-production. Masking, tracking or exploring a visual treatment may become less laborious. Yet computer-generated imagery, camera automation and virtual production are not all synonymous with AI. Nor is there a single ‘perfect’ colour grade: the choice belongs to the film’s visual language.

Lowering the cost of some tasks could give smaller teams room to experiment. It does not remove the work of directing, maintaining continuity, editing sound or finishing a film. Access to tools is also different from access to funding and distribution. The opportunity for independent production is real as a possibility, not a guarantee that every project will find its audience.

Recommendation systems influence which stories become visible. They can help someone discover a film, or repeatedly offer variations of what that person already watches. Interactive narratives raise a separate possibility: allowing viewers to make choices within a story. Branching narratives do not require AI, and more choice does not automatically create a more meaningful experience.

These changes affect work as well as images. Some repetitive tasks may shrink while supervision, correction and coordination become more important. Training should help people understand the tools and their limits without treating everyone as a machine-learning engineer. Storytelling, cinematography, performance and editorial judgement remain central skills.

The cultural risk is repetition at a larger scale. Systems built around widely distributed films may reproduce their conventions and leave less room for local rhythms, humour and visual traditions. Yet more affordable production could also help regional voices. Which tendency prevails depends on the choices of creators, commissioners and platforms, not on the technology alone.

Translation, subtitling and dubbing could make work accessible to more audiences, but language is more than literal meaning. A joke, accent or pause may carry the identity of a character. Human review and the approval of substantial changes matter if localisation is to open a work to another audience without flattening it.

My question is therefore not whether AI will make cinema more sophisticated. It is who gains the ability to tell a story, who retains control over its form, and which stories we are encouraged to see. A richer visual culture would require deliberate choices about all three.