AI Versus The Human Condition in filmmaking

This piece was born from watching the trailer for Dreams of Violets, the fully AI-generated feature programmed at Tribeca, and from my own reaction to that trailer alongside other examples of AI-generated film and video content I have seen over the past year.

AI is already part of the creative process. It is already being used for research, development, image generation, pitch materials, previs, budgeting, marketing, post-production support and countless other tasks. Pretending otherwise is pointless. But there is a very different question sitting underneath all of this: can AI actually replace the human centre of cinema?

Right now, my answer is NO.

Not because the technology is unimpressive. It is often astonishing. Not because it will not improve. It obviously will. But because cinema is not simply the arrangement of images, music, dialogue and atmosphere. Cinema is a contract of belief between the audience and the characters on screen. That contract depends on presence, continuity, emotional truth and the feeling that a human being is living through something in real time.

That is where AI-generated video still breaks.

The Dreams of Violets Moment

On paper, Dreams of Violets is a landmark. A 75-minute, fully AI-generated, live-action-style docudrama accepted into a major film festival. It was reportedly made quickly and cheaply, using AI tools instead of a traditional cast, crew, sets and physical production infrastructure. The film is inspired by political events in Iran and has been framed by some as the kind of story that perhaps could not have been made safely through conventional means.

That is a serious argument, and it deserves to be taken seriously.

There are places in the world where filming is dangerous. There are political realities, security risks and budgetary constraints that prevent stories from being told. If AI allows certain filmmakers to visualize events that would otherwise remain unseen, that is not something to dismiss out of hand. But the trailer also exposes the central problem..

The images may be cinematic. The subject matter may be urgent. The intention may be sincere. Yet the audience reaction has been sharply divided, and in many cases openly hostile. Why? Because the issue is not simply whether AI can generate images that look like cinema. The issue is whether those images feel inhabited.

At the moment, they mostly don’t.

Uncanny Valley Is Not Just a Visual Problem

Most people talk about the uncanny valley as if it is purely about faces looking slightly wrong. That is part of it, but it is not the whole issue. The deeper problem is performance..

AI video can now create faces that, in a still image, may look impressively close to photoreal. But film is not a still image. Film is time, movement, behaviour and emotional progression. It is the pause before a line. The blink that comes half a second too late. The breath an actor takes before deciding whether to speak. The body tightening before an emotional reveal. The eyes carrying information the dialogue does not.

That is where AI still struggles.

In dramatic cinema, audiences are incredibly sensitive to the smallest human signals. We may not consciously analyse them, but we feel them. We know when grief is being performed truthfully and when it is being approximated. We know when fear travels through a body and when a face is merely arranged to look afraid. We know when an actor is listening, thinking, withholding, lying or breaking.

AI can imitate the surface of those things, but it does not yet convincingly sustain them.

That is why so much AI-generated drama currently feels hollow. The characters move, but they do not seem to live. Their faces change, but we do not feel an inner life behind the change. Their eyes may look human, but they often fail to carry intention.

And cinema lives or dies on intention.

Character Drift Destroys Emotional Trust

One of the biggest problems in AI-generated video is consistency over time.

A character may look one way in one shot, then subtly different in the next. Their face shifts. Their proportions change. Their clothing alters. Their hands behave strangely. Their mouth movements almost match the dialogue, but not quite. Backgrounds smear or mutate. Objects appear, disappear or change shape. The physical world feels unstable.

These may sound like technical problems, but they create real world emotional problems.

An audience connects to a character partly because that character feels continuous. We believe we are watching the same person carry experience from one moment to the next. That continuity is the foundation of empathy.

If the face drifts, the body drifts, the environment drifts and the emotional register drifts, the audience may not be able to identify each error specifically, but they sense the instability. Something feels wrong. And once that happens, the viewer stops leaning in and starts evaluating the image as a technology demonstration.

That is a fatal flaw when trying to create human connection.

Mood-Board Cinema Versus Lived Drama

There is another issue that I think is especially relevant to producers and filmmakers.

A lot of AI-generated content currently feels like a mood board pretending to be a film.

That’s not meant as a cheap insult. In fact, AI is extremely good at mood boards. It can create tone, texture, lighting references, genre atmosphere, production design concepts and emotionally suggestive imagery at remarkable speed. For development, pitch decks, look books and investor materials, that can be genuinely useful. But atmosphere is not drama.

Drama requires causality. A character wants something. Something blocks them. They make choices. Those choices have consequences. The pressure escalates. The character changes, fails, sacrifices, adapts or reveals themselves. The audience understands not only what is happening, but why it matters.

A trailer can appear powerful because it compresses mood, music, imagery and fragments of emotion into a short burst. But a feature film has nowhere to hide. Over 75 or 90 minutes, the audience needs continuity, escalation, behavioural truth, character psychology and dramatic payoff.

The longer the format, the more exposed the weakness becomes.

The Job Displacement Question

There is no honest discussion about AI in film without discussing labour.

The fear across the industry is not theoretical. Writers, actors, concept artists, storyboard artists, VFX artists, editors, voice performers, background actors, composers, production designers, post-production workers and many entry-level crew are all looking at this technology and asking the same question: where does this leave us?

That concern and the anxiety it creates is legitimate.

If one person can create something that resembles a feature film from a laptop, the economic implications are obvious. Producers, financiers and studios will look at cost. They will look at speed. They will look at scalability. They will ask what can be done with fewer people, fewer shoot days, fewer locations and fewer logistical headaches.

The industry has always chased efficiency, but there is a difference between removing friction and removing authorship. There is a difference between using AI to support a filmmaker and using AI to eliminate the ecosystem that makes filmmaking possible. The latter may look attractive on a spreadsheet, but it carries serious creative, ethical, legal and commercial risk.

Actors are not just visual assets. Writers are not just prompt suppliers. Directors are not just image selectors. Crew are not just cost centres. They are the human intelligence of the production.

Strip too much of that away and you might save the bottom line, but you also reduce the value of human capital, which leads me to...

The Acting Problem

The acting question is probably the most sensitive and the most important.

Can AI create a synthetic performance? Technically, yes. Can it create an emotionally convincing dramatic performance sustained across an entire feature? Not yet, at least not in a way that consistently competes with a real actor.

Acting is not just expression. It is interpretation. It is choice. It is contradiction. It is the actor bringing experience, instinct, vulnerability and unpredictability into the scene. It is the difference between what a character says and what they mean. It is the body betraying the dialogue. It is chemistry between performers. It is the accident that becomes the best moment in the take.

AI can produce an image of sadness. It cannot yet give you a person trying not to be sad.

That distinction is everything.

For horror, thriller, drama and social-impact storytelling, where Skinny Dipped Films lives creatively, performance is not optional. It is the delivery mechanism for fear, guilt, grief, moral conflict, trauma, dread and catharsis. If the audience does not believe the person, they will not believe the story and that erodes trust.

Disclosure and Audience Trust

Audiences may tolerate AI when it is used transparently and tastefully. They are far less forgiving when they feel deceived. If a film, trailer, advertisement or performance is AI-generated or materially AI-altered, and that fact is hidden or minimized, the backlash can be severe.

This is partly an ethical issue and partly commercial.

The audience isn’t stupid. They can sense when something is off. If they later discover that AI was used in a way they feel was concealed, the problem becomes less about the technology and more about trust.

For producers, this matters. For distributors, it matters even more. For financiers, it should matter immediately.

AI use will increasingly become part of chain of title, E&O, guild compliance, performer consent, marketing positioning and distributor risk assessment. The question will not simply be “did you use AI?” It will be: what did you use, how did you use it, whose work trained it, whose likeness was involved, what permissions exist, and how clearly was it disclosed?

That is where the legal and commercial complexity begins.

So What’s The Bottom Line?

From a financing perspective, the current AI conversation needs to be separated into two very different categories.

The first is AI as a tool.

This is already real. AI can be useful for development analysis, pitch materials, visual references, previs, storyboarding, temp marketing assets, VFX planning, environment extensions, background elements, language support, post-production workflows and cost reduction. Used properly, it can help independent producers move faster, present more professionally and stretch limited resources.

That is the sensible near-term opportunity.

The second category is AI as the replacement for the entire production stack.

That is where I think the market is much less proven and a place I hope we don’t get to. In fact I don’t believe we will get to.

A fully AI-generated feature may attract festival attention because it is novel, provocative and culturally relevant. But festival curiosity is not the same as audience demand. It is not the same as theatrical viability. It is not the same as a distributor writing a meaningful cheque. It is not the same as talent wanting to attach themselves to the next one. And it is certainly not the same as a bankable model for a slate.

Right now, fully synthetic film still carries significant audience risk, guild risk, legal risk, reputational risk and distribution risk.

That does not mean it will always be that way. The technology will improve. The visual gap will narrow. Some filmmakers will use AI brilliantly. Hybrid workflows will become standard. Younger audiences may develop a different tolerance for synthetic imagery. New genres may emerge that are native to AI rather than imitations of traditional cinema.

But for the foreseeable future, the commercially credible model is likely to be hybrid: human-written, human-directed, human-performed stories, with AI used intelligently around the edges to enhance production value, reduce costs and expand creative possibilities.

In other words, AI as part of the filmmaking toolkit, not as a wholesale substitute for filmmaking itself.

Where I Land

I am not anti-AI. That would be hypocritical. I use it. I value it. I think it is already changing the business, and I think independent producers who ignore it will be at a disadvantage.

But I am also not persuaded by the idea that AI-generated video, in its current form, has solved cinema and I don’t believe it ever will.. my mouth to gods ear, lol

It has solved some image problems. It has solved some speed problems. It has solved some access problems. It may even solve some production problems. But it has not yet solved the human problem.

And the human problem is cinema.

Cinema works because audiences believe in people. Not just faces. Not just voices. People. Characters who seem to think, feel, remember, suffer, desire and change. The Human Condition.

Until AI can sustain that kind of emotional truth, it will remain an extraordinary tool, a disruptive force, but not a replacement for actors, filmmakers or the human instinct at the centre of storytelling.