The Sins of 3D #3: The 3D Conversion Myth
There is probably no label that has damaged the perception of stereoscopic conversion more than these two words:
Converted to 3D.
For many 3D enthusiasts, there has been an almost automatic assumption for years:
Shot in native 3D = real 3D = good.
Converted to 3D = fake 3D = bad.
It sounds logical. If two cameras captured the scene from two slightly different positions, just like our eyes, surely that must produce better 3D than creating the second viewpoint afterward…
Except modern filmmaking doesn’t work like that.
And neither does good stereoscopic 3D.
The distinction between native and converted 3D is far less clear than it appears. Movies advertised as being shot in 3D can contain converted shots. Huge portions of modern movies never existed in front of a physical camera at all. And even footage captured with two cameras may need extensive work before those two images are suitable for a theatrical 3D presentation.
Meanwhile, professional conversion gives filmmakers something that physical cameras cannot provide as easily: the ability to design depth after the shot has been captured.
So perhaps we’ve been asking the wrong question all along.
Instead of asking whether a movie was shot or converted in 3D, we should be asking:
Is the 3D actually good?
Welcome to The Sins of 3D #3.
And to The 3D Conversion Myth.
“Native 3D” doesn’t mean everything was filmed with two cameras
Let’s start with something obvious that is surprisingly easy to forget.
Think about a modern superhero or science-fiction movie.
A spaceship flies over an alien city. A planet fills the background. Hundreds of spacecraft fight around it. Buildings collapse. Creatures run between explosions.
Which 3D camera filmed that?
None.
The spaceship doesn’t exist. The planet doesn’t exist. The city might not exist. Perhaps most of what you’re looking at doesn’t exist outside a computer.
Even movies genuinely produced using stereoscopic cameras contain visual effects, computer-generated characters, environments and entire shots that were never photographed by a pair of physical lenses.
The more CGI there is in a movie, the less useful it becomes to imagine “native 3D” as simply two cameras recording everything you eventually see on the screen.
Computer-generated imagery can, of course, be rendered stereoscopically from virtual cameras. That’s genuine stereoscopic imagery too. But no physical stereo camera photographed it.
This becomes particularly important when discussing movies such as Avatar. The Avatar movies are landmarks in native stereoscopic filmmaking, but Pandora isn’t a physical place that James Cameron simply pointed two cameras at.
Modern filmmaking is a mixture of photography, virtual cameras, performance capture, CGI, compositing and enormous amounts of post-production.
The finished stereoscopic image is what matters.
Not whether two physical lenses happened to be present at the beginning of the process.
Even native 3D movies use conversion
Here’s where the distinction becomes even more interesting.
The original Avatar is probably the movie most closely associated with the modern era of native 3D cinematography.
And even Avatar needed converted shots.
Barry Sandrew, founder of Legend3D, stated that approximately 40 shots in Avatar required 2d-to-3D conversion. The same source describes conversion being used in other productions shot with stereoscopic camera systems and notes that converted and camera-captured shots could be intercut without audiences being able to identify which was which.
That alone should make us reconsider the idea that a movie must belong exclusively to one category.
Another striking example is Transformers: Dark of the Moon. Despite its association with native stereo camera rigs, Legend3D said it ultimately converted more than half of the movie — around 78 minutes — including complex material involving major visual-effects work.
So what is it?
Native?
Converted?
Both?
Perhaps that’s the point.
A filmmaker doesn’t need to be loyal to a capture method. A filmmaker needs to deliver the best final images possible.
Two cameras also mean twice the opportunity for something to go wrong
Shooting stereoscopically creates a problem that doesn’t exist in the same way with a conventional camera:
you don’t need one good image.
You need two good images that work perfectly together.
The cameras need to remain correctly aligned. Focus needs to match. Optical differences need to be controlled. Reflections and polarization can differ between the two views. Vibration or tiny geometric discrepancies that might be irrelevant in a normal movie can become visible or uncomfortable when your brain tries to fuse both images.
Vertical misalignment and focus mismatches are among the known problems that stereoscopic camera rigs can introduce.
And filmmaking doesn’t happen in a laboratory.
Cameras move. Actors move. Things explode. Water splashes. Smoke fills the frame. Reflections appear. Equipment vibrates. Something unexpectedly blocks one camera but not the other.
Sometimes you don’t discover every problem until post-production.
And here’s the important part:
Nobody is going to keep a bad stereoscopic shot in an expensive movie just so they can claim it was “real 3D.”
If an expensive action shot is fantastic in one view but the stereo pair is unusable, you don’t necessarily rebuild the set, bring back the actors and spend another fortune repeating it.
You fix it.
And one possible solution is to use the usable image and reconstruct the stereoscopic version in post-production.
In other words: conversion.
The more difficult the shot, the less useful the label becomes
This is especially relevant to action scenes.
Ironically, these are often exactly the scenes where audiences most want spectacular 3D — and where capturing perfect stereoscopic imagery can become more complicated.
Now add CGI.
Then compositing.
Then digital doubles.
Then particles.
Then background replacements.
Then objects that didn’t exist during photography.
At what exact point do we stop calling that image “native”?
It’s increasingly the wrong question.
A modern movie frame isn’t necessarily a photograph. It can be a construction assembled from many different sources.
And the stereoscopic version of that frame is another part of that construction.
Live concerts make the problem even more obvious
Imagine filming a live concert in 3D.
The artist isn’t going to perform a song again because Camera 7 had a stereoscopic problem during the best moment.
Even in normal 2d live production, footage gets rejected. A camera loses focus. Someone walks in front of it. There’s too much movement. The framing isn’t good. Another camera simply captured the moment better.
Stereoscopic production adds another requirement: both views have to work together.
That gives the production additional ways for otherwise useful footage to become unsuitable.
If the best image of an unrepeatable moment exists only as a usable 2d image, what would we prefer?
Lose the best shot simply to preserve a “100% native” label?
Or convert it properly and keep the best possible edit?
For us, the answer is obvious.
The movie — or concert — should serve the viewer.
The viewer shouldn’t have to serve the camera technology.
Native cameras have a depth budget
There is another limitation of physical stereoscopic photography that is rarely mentioned when people use “native 3D” as a quality label.
A stereo camera doesn’t have unlimited depth available to distribute anywhere it wants.
There is effectively a depth budget.
The amount of stereoscopic disparity that remains comfortable depends on the geometry of the shot, the camera setup and where objects appear in relation to the viewer.
This creates two opposite problems.
Put something very far away and the difference between the left- and right-eye views becomes extremely small. Natural stereoscopic depth gradually disappears with distance.
But bring something excessively close to the cameras and you can create the opposite problem: the difference between the two views can become so large that the image becomes uncomfortable — or even difficult for the viewer’s eyes to fuse correctly.
So a stereographer cannot simply turn the 3D up to maximum and make everything from an actor’s nose to a mountain kilometres away display enormous depth simultaneously.
Something has to give.
Sometimes you have to choose where the depth goes
Imagine a shot with a character extremely close to the camera and a spectacular landscape extending kilometres behind them.
The nearby character may already consume a significant part of the comfortable stereoscopic range.
Configure the cameras to give the distant environment stronger disparity and you may make the foreground uncomfortable.
Keep the foreground comfortable and the distant landscape may end up with very little perceptible stereoscopic separation.
That’s not necessarily bad stereography.
That’s geometry.
And it demonstrates something extremely important:
Even if a movie is shot with stereoscopic cameras, the word “native” tells you almost nothing about how strong its 3D will actually be.
The filmmakers still have to decide how the available depth is distributed.
They can create restrained stereo.
They can create stronger stereo.
They can prioritize foreground objects.
They can prioritize the overall environment.
They can deliberately avoid aggressive negative parallax.
They can make different choices from shot to shot.
Two movies can therefore both be “100% native 3D” and provide completely different stereoscopic experiences.
The production label doesn’t tell you which one you’ll prefer.
But isn’t native 3D more realistic?
Yes.
And sometimes that’s precisely its limitation.
This may sound almost sacrilegious among stereoscopic purists, but consider what movies actually are.
Cinema isn’t reality.
If movies had to reproduce reality faithfully, explosions wouldn’t erupt at every opportunity. Lighting wouldn’t always make actors look perfect. Bullets wouldn’t conveniently miss protagonists. Landscapes wouldn’t be digitally enhanced until they look more spectacular than the real locations.
And you certainly wouldn’t hear enormous explosions in outer space.
We accept all of those things because cinema isn’t supposed to be a scientific reproduction of reality.
It is supposed to create an experience.
So why should stereoscopic depth be the one cinematic tool that isn’t allowed to exaggerate?
Real depth can actually be rather boring
Our binocular vision has physical limitations.
Stereopsis is extremely powerful at relatively close distances. But as objects become farther away, the angular difference between what our two eyes see becomes progressively smaller.
Put a huge spaceship several kilometres away and, despite its enormous physical volume, binocular disparity isn’t going to give you the spectacular stereoscopic volume you might expect from a science-fiction movie.
Even at much shorter distances, people can only see depth up to 10 meters away (0,01 Kilometers, or 0,0062 miles), some people even need objects much nearer than 10 meters to start appreciating some depth.
Now imagine a gigantic planet behind that spaceship.
Realistic stereoscopy says that all the frame -including the spaceship- should look totally flat.
Cinematic stereoscopy doesn’t necessarily have to.
A filmmaker can decide that the audience should perceive the volume of that spaceship. Or that mountains should separate beautifully into layers. Or that a planet in the background should feel spherical rather than like a flat disc pasted onto the sky.
Is that realistic?
Perhaps not.
But neither is the planet.
And it may produce a much more satisfying 3D experience.
The contradiction among 3D enthusiasts
This creates a rather amusing contradiction.
Some of the same enthusiasts who demand “real native 3D” are also among the first to complain when a movie follows natural stereoscopic depth too conservatively:
“The backgrounds are flat.”
“There’s hardly any depth.”
“Nothing comes out of the screen.”
“That shot was such a missed opportunity.”
But sometimes the supposedly disappointing result is closer to what our eyes would actually perceive at those distances.
You can’t simultaneously demand absolute optical realism and complain that most scenes of a movie don’t have enough stereoscopic separation.
At some point, we have to decide what we actually want from 3D cinema.
For us, stereoscopic 3D isn’t merely a simulation of standing in front of the actors.
It is a creative visual medium.
This is particularly noticeable when comparing different approaches to stereoscopic filmmaking.
The Avatar movies can produce extraordinarily convincing stereoscopy, particularly with characters, creatures, vegetation, particles and objects at relatively close distances. There is an exceptionally natural feeling to much of their spatial presentation.
But move into certain wider shots and landscapes and the depth can become considerably more restrained.
That’s physically logical.
It can also be disappointing if what you enjoy is strong, clearly perceptible stereoscopic depth throughout the entire image.
Now compare that experience with many of Marvel’s professionally converted 3D releases.
A face can have depth.
The room behind it can have depth.
The city outside can have depth.
A spacecraft can have depth.
The planet behind that spacecraft can have depth.
The stereoscopic presence can remain deliberately perceptible across an enormous range of apparent distances.
This isn’t an objective declaration that one philosophy is universally better than the other. Some viewers prefer naturalistic stereo; others enjoy a stronger presentation.
But it demonstrates something important:
The most realistic 3D isn’t necessarily the most enjoyable 3D.
And conversion gives stereographers enormous freedom to decide where that depth should exist.
With conversion, depth becomes part of the filmmaking
This is the fundamental advantage of professional conversion that is so often ignored.
You’re not merely trying to recover what two physical cameras would have seen.
You can design the stereoscopic experience.
Return to our character standing close to the camera with that spectacular landscape stretching far behind them.
A stereographer working in post-production isn’t completely constrained by what a physical stereo rig would have captured at that exact moment.
The character can retain convincing volume while the landscape is separated into perceptible layers. Different objects can occupy carefully designed positions in stereoscopic space. The depth can be adjusted to complement what the director wants the viewer to notice.
The stereographer can decide that a mountain should feel farther behind another mountain.
That a spacecraft needs more volume.
That a distant city should visibly extend toward the horizon.
That a planet should feel spherical.
Or, conversely, that a particular scene should deliberately become flatter because the audience needs to concentrate on one character.
This is an important distinction.
Good conversion isn’t about maximizing depth everywhere. It’s about controlling depth everywhere.
That’s not necessarily a failure to reproduce reality.
It’s stereoscopic cinematography.
Filmmakers already manipulate focus, perspective, focal length, lighting, exposure, color, time, motion and sound.
Depth can be another creative variable.
They may initially plan the 3D conversion—or set up the 3D cameras—with certain depth values in mind, only to realize later that a particular scene would work much better with a completely different approach to depth. With native 3D shooting, however, you cannot simply redefine the captured depth once the footage has been shot—unless you discard the original 3D information altogether and perform a new conversion from the 2d source.
Stronger doesn’t automatically mean better either
There is an important qualification to all of this.
If restrained 3D isn’t automatically better because it’s more natural, extreme 3D isn’t automatically better simply because there’s more of it.
Push disparities too far and the experience can become uncomfortable.
Use excessive pop-out constantly and it can become distracting rather than exciting.
Change the depth dramatically between consecutive shots without considering how the viewer’s eyes have to adapt and the result can become tiring.
A good stereoscopic presentation therefore isn’t simply one with the biggest possible numbers.
For us, great 3D should be visible, dimensional, consistent, comfortable and intentional.
We want to perceive it.
We want characters to have volume.
We want environments to extend convincingly behind the screen.
We want foreground, middle ground and background to occupy meaningful positions in space.
We enjoy objects coming toward or beyond the screen when the scene benefits from it.
But above all, the stereography should feel controlled.
That’s what separates strong 3D from simply excessive 3D.
Professional conversion isn’t pressing a “Make 3D” button
And this is where we reach one of the biggest misunderstandings surrounding converted movies.
A high-end theatrical conversion isn’t simply a computer automatically guessing a depth map and generating a second image.
Professional conversion has traditionally involved large teams of artists separating objects, defining geometry and depth, handling occlusions, reconstructing areas hidden in the original view, correcting difficult edges and dealing with things such as hair, transparency, reflections and particles.
The stereoscopic result can then be reviewed and adjusted.
Shot by shot.
And where necessary, frame by frame.
Conversion is actually an extremely labor-intensive process requiring considerable stereographic expertise when performed at a high level. It also provides substantial creative freedom — which means there are more decisions to make, not fewer.
The second viewpoint may be artificial. The love & work put into so many 3D-converted movies is not.
The craftsmanship isn’t. There are tens of 3D converted movies with better 3D than Avatar.
One rushed decision gave 3D conversion a terrible reputation
Then came Clash of the Titans.
Released in 2010, immediately after Avatar had created enormous excitement around digital 3D, Clash of the Titans became perhaps the most infamous example of a rushed post-conversion.
The decision to release it in 3D came late in production, amid an already compressed post-production schedule. Contemporary coverage described the 3D conversion as a late addition to a project whose visual-effects team was already working under substantial time pressure.
Its reception helped create an association that stereoscopic cinema has struggled with ever since:
conversion = bad 3D.
But that’s the wrong lesson.
The lesson should have been:
Bad conversion = bad 3D.
A badly photographed native stereo movie isn’t proof that stereoscopic cameras don’t work. A big budget movie can also have horrible CGI effects, We should say that CGI looks bad because on one example they did wrong? Absolutely not!
Likewise, a rushed conversion isn’t proof that conversion doesn’t work.
Unfortunately, labels are easier to remember than production circumstances.
And now, sixteen years later, we may be creating a new version of the same problem.
Anyone can convert a movie to 3D now
Something remarkable has happened in the last few years.
You don’t need a Hollywood visual-effects studio anymore to turn 2d video into stereoscopic 3D.
AI-powered applications such as Owl3D can analyze ordinary video, estimate depth and generate stereoscopic left- and right-eye views automatically. Owl3D itself describes a process in which AI generates a depth map and uses it to create the displaced viewpoints required for stereoscopic viewing.
As 3D enthusiasts, we think that’s fantastic.
People can experiment with their own footage. They can experience material in 3D that would otherwise never receive a stereoscopic release. It demonstrates both how far depth-estimation technology has progressed and how much demand still exists for more 3D content.
But it introduces a new problem.
An automatic 3D conversion isn’t the same thing as a professionally authored 3D conversion. Not at all.
AI can estimate depth. That doesn’t make it a stereographer
Modern AI can produce remarkably convincing depth maps.
But depth estimation and stereography aren’t the same task.
Even Owl3D’s own technical material acknowledges one of the fundamental difficulties: depth needs to remain temporally consistent between frames. If the estimated depth of an object changes from one frame to another, the resulting stereoscopic image can become unstable. The software therefore includes specific processing designed to improve temporal consistency.
And that’s only one aspect of the problem.
What should be behind the screen?
What should come forward?
How strong should the depth be in this particular shot?
Should the next shot have the same depth?
Should that distant landscape be exaggerated?
Should the character be separated more strongly from the background?
Should the depth gradually change during the shot?
Is the director trying to make us look at the foreground or the background?
An algorithm can estimate where things are.
A stereographer decides how we should experience them.
That’s a very different job.
AI isn’t the problem either
This distinction is important because we don’t want to create another false argument.
This isn’t:
AI conversion = bad.
Just as the original argument shouldn’t be:
converted = bad.
AI can become an extraordinarily powerful tool for professional stereoscopic production.
The important distinction is between automatic conversion and professionally supervised conversion.
An AI system can save enormous amounts of work by generating excellent initial depth information. Artists can then refine it, correct errors and, most importantly, make creative decisions about the stereoscopic presentation.
The image above perfectly demonstrates that even with automated conversions, there are a lot of choices about depth for every element of every frame.
A huge team spending months following a carefully planned approach to a movie’s 3D and ensuring consistency throughout is highly unlikely to produce a film with bad 3D.
That’s very different from feeding an entire movie into an application and accepting whatever comes out.
Both files might subsequently be described online using exactly the same words:
“3D conversion.”
Yet the production processes — and potentially the results — could hardly be more different.
The new danger: someone’s first 3D movie might be an amateur conversion
And this is where the subject belongs in The Sins of 3D.
Imagine someone who has just bought their first device capable of showing stereoscopic movies.
They search online for 3D content.
Instead of an official stereoscopic release, they find an automatically converted version produced by another enthusiast.
Perhaps some scenes look surprisingly good.
Others don’t.
Depth changes unpredictably. An object is interpreted incorrectly. Fine details cause problems. One shot has strong depth and the next barely any. Something that should be behind something else appears in the wrong spatial relationship.
The newcomer doesn’t know anything about the conversion process.
They don’t think:
“This particular depth-estimation model made a mistake.”
They think:
“So this is what converted 3D looks like.”
Or worse:
“So this is what 3D looks like.”
And another potential 3D enthusiast walks away believing that the format itself was the problem.
Amateur 3D conversions are not the problem. Experimenting, learning, converting movies out of passion, and trying to make more content available in 3D help keep the format alive. The problem begins when those conversions are distributed too casually, without considering whether they are consistent and polished enough to become someone else’s first—or only—impression of 3D.
Enthusiasm alone does not guarantee a good viewing experience. A conversion may contain spectacular moments and still be plagued by inconsistencies, uncomfortable depth, distracting artifacts, or shots where the illusion suddenly falls apart. The person who created it knows that it is an amateur conversion and understands the limitations of the process. The person who receives it may not. They simply watch a movie in 3D, have a poor experience, and conclude that 3D itself is gimmicky, uncomfortable, cheap, or simply not worth the trouble.
That is the real danger of distributing amateur conversions without sufficient quality control. You are not only sharing your work; whether you intend to or not, you may also be representing the format to someone who has no other reference.
3D has spent more than fifteen years struggling with a reputation for inconsistency, gimmickry, and poor execution. Carelessly distributing unfinished or uneven conversions risks reinforcing exactly the perception that has kept the format trapped in that cycle. Good intentions can therefore have the opposite effect: instead of creating another 3D enthusiast, a mediocre experience may convince someone that 3D was never good in the first place.
So the problem is not amateur 3D conversion. Quite the opposite: experimentation and passionate creators can be enormously valuable to the format. The problem is treating every successful conversion as automatically ready for distribution. Creating something for yourself is experimentation; putting it into other people’s hands comes with a different responsibility. If we want 3D to escape the reputation for cheapness that has surrounded it for so long, the standard for what we distribute should be higher than simply “it works.”
That’s another Sin of 3D.
If we want more professionally produced 3D, we also need to support it when it is actually offered. Piracy may have acted as a Trojan horse, helping 3D reach audiences that might otherwise never have discovered it, but circulation alone cannot sustain professional production. Supporting official 3D releases sends a much stronger message: that there is an audience willing to pay for quality. And the stronger that message becomes, the easier it is to justify not only new 3D releases, but hopefully professional conversions of older films as well. The goal should not be to discourage amateur conversions, but to help build a future in which we no longer have to depend on them.
There’s an extraordinary irony here
You may even encounter 3D enthusiasts criticizing a Hollywood movie because:
“It wasn’t shot in real 3D. It was converted.”
And then those same enthusiasts may use consumer AI software to automatically convert a movie that has no official 3D release.
There’s absolutely nothing wrong with doing that.
But think about the contradiction.
The Hollywood conversion they rejected might have involved months of work by professional artists, shot-by-shot stereographic decisions, manual corrections and frame-by-frame intervention.
Their home conversions have been generated automatically from AI-estimated depth, resulting in shaky images, halos, and artifacts throughout. The quality can vary dramatically from shot to shot—or even within the very same shot: what looks perfectly acceptable one moment can suddenly fall apart with nothing more than a slight change in the image.
There is a strange contradiction in refusing to pay for a professionally converted 3D movie, where serious flaws are rare, while readily accepting an inferior home conversion—one generated by an algorithm the enthusiast has little real control over, with artifacts and visual discomfort appearing throughout.
Yet both are dismissed with the same word:
Converted.
If anything demonstrates how meaningless that word has become as a measurement of quality, this does.
We shouldn’t discourage amateur conversion
Quite the opposite.
The fact that enthusiasts are willing to spend their own computing time converting movies tells us something important:
There isn’t enough professional 3D content available.
People want to see movies in 3D that studios have never released in 3D.
Automatic conversion is filling a vacuum.
And today’s tools can produce results that would have seemed extraordinary on a home computer not very long ago.
The problem isn’t that these tools exist.
The problem comes when an automatically generated conversion is mistaken for evidence of what professional stereoscopic conversion can achieve.
We need better language and better understanding.
Perhaps we should stop treating “converted” as a quality category altogether.
There are good and bad conversions — just as there is good and bad native 3D, photography, visual effects and so on.
None of this means every professionally converted movie has great 3D.
Of course not.
Some conversions are conservative. Some have less depth than we would like. Some make creative decisions we disagree with. Some simply aren’t particularly impressive.
Productions recorded with 3D cameras can disappoint too.
That’s precisely the point.
The production method doesn’t guarantee the result.
A camera is a tool.
Conversion is a tool.
CGI is a tool.
AI is becoming another tool.
What matters is how those tools are used.
The best 3D is the 3D you can actually see
So how should we judge a 3D movie?
Not by reading a production label.
Watch it.
How strong is the depth?
How much volume do characters and objects have?
How clearly can you perceive the separation between foreground, middle ground and background?
Do large environments make effective use of stereoscopic space?
Does the depth remain consistent from scene to scene?
Are moments of negative parallax used effectively?
Is the presentation comfortable enough to enjoy for the duration of the movie?
Does the stereography contribute something meaningful that you would lose by watching the 2d version?
And most importantly:
Is it enjoyable?
Those questions tell us far more about the quality of a 3D presentation than knowing whether two physical cameras were used during principal photography.
Stop asking whether it’s native or converted
Perhaps it’s finally time to retire one of the simplest assumptions of the digital 3D era.
Don’t ask:
“Was it shot in real 3D?”
Ask:
“How good is the 3D?”
Because native 3D and converted 3D aren’t enemies.
They’re tools available to filmmakers trying to achieve the same thing:
Two images that convince our brains that there is a world behind — and sometimes in front of — the screen.
A great native production can capture wonderfully natural volume and genuine binocular information directly from the scene.
A great conversion can create depth that would be difficult, impractical or sometimes impossible to obtain with a physical stereoscopic camera setup. It can independently control different parts of the image. It can integrate live action with digital objects that never existed in front of any camera. And it can rescue footage where the original stereoscopic capture simply didn’t work.
And increasingly, modern productions can use both.
We don’t have a conversion problem.
If the result is spectacular, consistent, and comfortable, does it really matter whether the second image was captured through another lens, generated by a virtual camera, crafted through months of painstaking work by stereoscopic artists, or created through a combination of all three?
And that will lead directly to the next chapter of this story.