WHY MOST AI FILMS STILL LOOK LIKE AI
By Nick Sharma · Founder & Creative Director, Rude X
There is a moment that happens quite often when I'm watching AI-generated video.
The first shot looks incredible.
The lighting is beautiful. The environment is detailed. The character looks convincing. The camera movement is smooth. For a second, you forget you're looking at something generated.
Then the next shot arrives.
And suddenly, you know.
You can feel it.
It looks like AI.
The strange thing is that nothing necessarily went catastrophically wrong. The face might be perfectly intact. The hands might have five fingers. The image might be sharp enough to make a cinema camera nervous.
Yet something feels off.
That's what interests me.
Because I don't think the biggest problem with AI filmmaking is that AI hasn't become realistic enough.
I think the bigger problem is that we're often asking it to create images that look cinematic without actually thinking cinematically.
And those are two very different things.
PHOTOREALISM ISN'T THE SAME AS REALISM
We've become obsessed with photorealism.
It's understandable. Early generative imagery often looked obviously synthetic, so realism became the benchmark. We wanted skin to look like skin. Hair to behave like hair. Fabric to have believable texture. Light to behave according to physical rules.
All of that matters.
But here's the uncomfortable part:
A photorealistic image can still feel completely unreal.
Think about a traditional film. Even when you're looking at something fantastical, you're rarely analysing every pixel and asking whether it looks real. Your brain accepts the world because the visual language is coherent.
The camera behaves like a camera.
Light has direction.
Objects have weight.
Characters occupy space.
Movement has intention.
The environment responds to what is happening inside it.
Everything belongs to the same reality.
AI often gives us individual moments that look convincing, but those moments don't necessarily belong to the same reality.
That's where the illusion starts to fall apart.
AI DOESN'T KNOW WHAT MATTERS
A cinematographer doesn't simply record everything that exists in front of the lens.
They decide what matters.
That's the entire game.
A frame can contain a hundred pieces of information, but the cinematographer decides where the audience's eye should go first.
Maybe it's the face.
Maybe it's a hand.
Maybe it's a doorway in the distance.
Maybe it's the silhouette of a character against a bright sky.
The rest of the image supports that decision.
Generative models, by contrast, are incredibly good at giving you more.
More texture. More detail. More atmosphere. More dramatic lighting. More visual complexity.
And sometimes more is exactly what ruins the shot.
I've generated frames that were technically spectacular and creatively useless.
Everything looked expensive.
Nothing looked important.
That distinction matters.
A cinematic image isn't a competition to see how much information we can squeeze into the frame.
Sometimes the most cinematic choice is removing information.
Letting a face fall into shadow.
Leaving negative space.
Allowing a background to disappear into darkness.
Holding a static composition while something subtle happens in the foreground.
Cinema is often about restraint.
AI tends to tempt us toward excess.
THE CAMERA IS MOVING BECAUSE IT CAN
This is one of the easiest ways to spot AI filmmaking.
The camera moves.
It moves beautifully.
It pushes in. It circles. It cranes upward. It flies through the environment. It moves around the subject in a perfect arc.
And you find yourself wondering:
Why is the camera doing that?
That's a filmmaking question, not a technology question.
Camera movement is supposed to communicate something.
A slow push can create intimacy.
A pullback can reveal isolation.
A lateral move can reveal information.
A handheld camera can create instability.
A locked-off shot can create tension.
A rapid movement can create energy.
But when every AI shot contains movement simply because movement looks impressive, the camera stops speaking.
It becomes decoration.
This is one of the reasons AI reels can feel exhausting. Every shot is trying to impress you.
There is no breathing room.
No visual hierarchy.
No rhythm.
Just movement.
More movement.
Then even more movement.
Eventually, the audience stops feeling anything.
THE LENS MATTERS EVEN WHEN THERE IS NO LENS
One of the strange things about AI cinematography is that you can describe a lens, a focal length, a depth of field and a camera system, but the model doesn't necessarily think about those things in the way a cinematographer does.
It approximates their visual consequences.
That's incredibly useful.
But there's a difference between saying, “make this look like a 50mm shot,” and understanding why a 50mm lens might be appropriate.
Lens choice changes how we experience space.
Wide lenses exaggerate proximity and depth. Longer lenses compress space and isolate subjects. A shallow depth of field can draw attention toward a face. A deeper focus can make the relationship between foreground and background important.
These aren't aesthetic filters.
They're storytelling decisions.
When AI filmmaking uses lens language as a vocabulary of visual adjectives rather than a tool for storytelling, the result can look impressive without feeling intentional.
The image is borrowing the appearance of cinematography without necessarily understanding its grammar.
LIGHTING ISN'T JUST “CINEMATIC LIGHTING”
Another favourite phrase in AI prompting is “cinematic lighting.”
It's become almost meaningless.
What does it actually mean?
Where is the light coming from?
Why is it coming from there?
How hard is it?
What colour is it?
What is motivated by the environment?
What's being hidden?
What's being revealed?
What happens to the character when they move?
Real cinematography has consequences.
A character walking past a bright window should interact with that light.
A subject moving beneath streetlights should experience changes in exposure and colour.
A cloud passing over the sun should affect the environment.
A practical light in the background should feel like it belongs to the world rather than being painted there because it looks pretty.
When AI lighting looks artificial, it isn't always because the light itself is unrealistic.
Sometimes it's because nothing else responds to it.
The light exists.
The world doesn't.
MOVEMENT REVEALS MORE THAN STILL IMAGES DO
AI image generation has become remarkably good at creating convincing still frames.
Video is much less forgiving.
The moment something moves, the illusion has to survive time.
A character can't just look correct.
They have to remain the same person from one moment to the next.
Clothing needs to maintain its relationship with the body.
Objects need to maintain scale.
Hair needs to behave consistently.
Light needs to continue making sense.
The environment needs to remain spatially coherent.
And the camera has to move through that space without quietly changing the geometry of the world every few frames.
This is why cinematic AI isn't simply about generating better frames.
It's about continuity.
A beautiful frame is easy to admire.
A believable sequence is much harder to construct.
CHARACTERS AREN'T JUST FACES
AI has become very good at generating attractive people.
That's not the same thing as creating believable characters.
A face is only one part of a performance.
Posture matters.
Eye direction matters.
Breathing matters.
Weight distribution matters.
Small gestures matter.
The speed of movement matters.
Even stillness matters.
One of the biggest differences between a generated person and a believable character is that a believable character appears to exist before and after the shot.
They have intention.
They aren't simply standing there looking beautiful.
This is particularly important in AI filmmaking because generative systems are very good at creating what might be called the appearance of performance.
The real challenge is creating behaviour.
And behaviour is storytelling.
THE EDIT CAN MAKE AI LOOK MORE LIKE AI
This one is often overlooked.
You can generate beautiful footage and still destroy it in the edit.
Why?
Because AI filmmakers often feel compelled to show every good shot they have.
I understand the temptation.
You spend hours generating something spectacular. You've finally got the shot. Of course you want to use it.
But a film doesn't care how difficult a shot was to generate.
It cares whether the shot belongs there.
Sometimes an AI sequence becomes more convincing simply by removing half of it.
Hold the shot longer.
Cut earlier.
Give the viewer time to understand the space.
Let one image contrast with the next.
Use a quiet shot after something chaotic.
Let the rhythm breathe.
Editing creates meaning between images.
That's something generation alone cannot solve.
THE BIGGEST AI TELL ISN'T A GLITCH
It's repetition.
The same type of camera movement.
The same shallow depth of field.
The same dramatic lighting.
The same perfect skin.
The same slow motion.
The same heroic poses.
The same impossible environments.
The same attempt at “epic.”
Eventually, AI filmmaking starts to develop its own visual accent.
And ironically, the more people try to make their work cinematic, the more similar it can become.
That's why I think one of the most important skills in AI filmmaking is learning when not to use the obvious choice.
Maybe the hero doesn't need to look heroic.
Maybe the camera shouldn't move.
Maybe the light should be ugly.
Maybe the world should be quiet.
Maybe the shot should be wide.
Maybe the subject should barely be visible.
Maybe the most interesting thing in the frame isn't the character at all.
Cinema has spent more than a century teaching us that visual storytelling doesn't require everything to be beautiful.
AI filmmaking needs to rediscover that lesson.
STOP ASKING AI TO MAKE IT CINEMATIC
This may be the most important thing I've learned.
Don't ask the machine to make something cinematic.
Make the creative decisions cinematic.
Decide what the audience needs to feel.
Decide what the shot needs to communicate.
Decide where the camera should be.
Decide what the light is doing.
Decide what deserves attention.
Decide how long the shot should last.
Decide what comes before it.
Decide what comes after it.
Then use AI to help you build the image that supports those decisions.
That changes the workflow completely.
Instead of asking:
“How do I make this look cinematic?”
you're asking:
“What does this shot need to accomplish?”
The first question produces an aesthetic.
The second produces a film.
AI ISN'T FAILING. WE'RE ASKING IT THE WRONG QUESTION.
I don't think the answer is to wait for AI models to become perfect.
They will get better.
Motion will improve. Consistency will improve. Physics will improve. Character performance will improve. Camera control will improve.
But even when the technology becomes extraordinarily capable, the fundamental problem won't disappear.
Someone still has to decide what to do with it.
That's the exciting part.
AI gives filmmakers access to an enormous visual possibility space. The challenge is no longer simply getting an image that looks good.
The challenge is choosing the image that means something.
That's where cinematography comes back into the conversation.
Because the future of AI filmmaking isn't going to belong solely to the people who know how to write the longest prompts.
It will belong to the people who understand why the shot exists in the first place.
And when that happens, maybe we'll stop talking about whether something was made with AI.
We'll just call it a film.
Nick Sharma
Founder & Creative Director, Rude X
Nick explores the intersection of filmmaking, generative AI, visual storytelling and creative technology, with a particular obsession for cinematic worlds that don't exist — yet.
Comments