You'll find here an open knowledge base concerning the main techniques for shooting in S3D, along with a couple of advices and good practices. The idea of this Wiki if to share as many info and knowledge as possible. Polemics and opinons are absolutely welcome on Convergence3D.net, but maybe more on the Forum rather than in these pages, where we'd like to try to summarize what is accepted by most stereographers.

The reproduction of the texts of this section is free, provided that its origin (“Stereo 3D Filmmaking: The Complete Interactive Course; convergence3d.net”) and its authors (“David Steiner and Cédric-Alexandre Saudinos, Parallell Cinéma”) are mentionned.

How to optimize the planning of a S3D shooting to speed up production?
Writing for S3D
Pre-production without a specialized software program
Pre-production with a specialized software program (FrameForge)

NOTE : These texts are based on the course by David Steiner and Cédric-Alexandre Saudinos “Stereo 3D Filmmaking : The Complete Interactive Course”, a DVD set with 14h of content, including 10h in S3D, which details, analyzes and illustrates theory of S3D and all of the practical techniques mentionned here. For more details, have a look here: convergence3d.net/stereo3d.php

Differences between the left and the right image of an object appear when the two images are superimposed. These differences are called parallaxes. A parallax is the distance between the left and the right image of the same object. It is expressed as a percentage of the total image width.

A parallax is POSITIVE when the left image of an object is on the left, and the right image is on the right. In this case, the object feels far, behind the screen plane. quand l’image de gauche d’un objet est à gauche, et son image de droite à droite. Exemple : +1%

There is NO PARALLAX when the two images of an object are perfectly superimposed. In that case, the object appears on the screen plan (= exactly at the distance of the porjection screen, not behind it or in front of it). That's 0% parallax.

A parallax is NEGATIVE when the left image of an object is on the RIGHT, and its right image is on the LEFT. In that case, the object feels close, in front of the screen, in “out-of-screen effect”. Exemple : -2%

Note : Parallaxes can also be expressed in pixels. In that case, 19 pixels are 1% of a 1920 pixels-wide image (that's Full HD). But, of course, 1% is 40 pixels on a 4K image and only 13 pixels for a 1280 pixels-wide image. To avoid confusion, we'll only speak in percentages in these texts - conversion is easy.

We already know there is a limit for the parallax of the farthest object. That's because the space between a human's eyes (called the “interocular”) is 2.5inches (6.5cm) on average.

So, if on screen, an object has 2.5inches/6.5cm between its left and right images, then my eyes have to look straight ahead: my left eye looks in front of himself, and my right eye looks also straight in front of himself. That means that they have to be exactly parallell in order to fuse the two images of that object. This gives me the feeling that this object is “at infinity”, the impression that this object is the farthest object I can see.

But if on screen, an object has more than 2.5inches/6.5cm between its left and right images (even if the distance between my eyes is still 6.5cm), then to fuse the images my left eye must look to its left; and my right eye has to look… on its right! My eyes are in divergence. This kind of situation isn't natural for the human eye - at best it is tiring, at worst it is painful and leads to headache and nausea.

We may infer that a postive on-screen offset of 6.5cm/2.5in between the two images of an object is the maximum we can accept. For that reason, the parallax that leads to an on-screen offset of 6.5cm/2.5in on a specific screen is often called the “artificial horizon”.

6.5cm/2.5in is an absolute value. It means that a given film is not necessarly adapted to any screen size. On a small screen, a 1m-wide screen (40in-wide), a certain offset will be, say, 1cm (0.4in): this is 1% of the width of the screen. But on a 13m-wide screen (500in-wide), the same 1% parallax will lead to an on-screen offset of 13cm (or 5in). And on a 26m-wide IMAX (1000in or 85ft), the on-screen offset for the same 1% parallax will be 26cm (10in): this leads to a very bad divergence! So: choosing the parallax of the farthest object according to the screen size we're aiming is very important.

We understand perfectly that a far parallax of 1% of the width of the frame is perfectly OK for a 6.5m screen (250in), because 1% is on that screen size 6.5cm/2.5in; but will be a bit disappointing on a smaller screen (a TV screen, for exemple), and problematic on a 26m screen!

We can deduce that:

The Artificial Horizon of a 6.5m/250in screen is +1% (the ”+” means it's a positive parallax).

The Artificial Horizon of a 13m/500in screen is +0.5%

The Artificial Horizon of a 26m/1000in screen is +0.25%

The small screen case is special: small screens are most of the time very close to the spectator. So big parallaxes can force the eyes to take quite significant angles. Experience shows that on those screens, most people are comfortable with an Artificial Horizon of +1.8% to +2.5% (for a film - for still images, one can push the Artificial Horizon further than that, because the audience does have as much time as they want to fuse the image) - and +2.5% is already considered to be quite a lot. When values are superior to that, most of the time the shot becomes too long to fuse to be used effectively in a movie.

So:

The Artificial Horizon of a small screen is about +2%.

Note that for negative parallaxes (they lead to out-of-screen effects), there is no real “negative artifical horizon”, since convergence is not as problematic as divergence.

But one should keep in mind that a -1% parallax for an object will give the impression that this object is at 1/2 the screen-to-spectator distance on a 6.5m/250in-wide screen; and at 2/3 of this distance on a 13m/500in-wide screen - a very powerful effect.

Shooting stricly parallel (for IMAX)
Shooting parallel with a post-production shift
Shooting convergent

NOTE : The Interaxial is set when the far parallax has already been set, either by shifting one image in comparison to the other in post-production; or by angulating a camera.

The Interaxial is the distance between the center of the lenses of the two cameras.

Even if this is a very common mistake, one should not make a confusion with the interocular, which is the distance between the optical axis of the eyes (6.5cm/2.5in on average). Interaxial and interocular are independent, since the interaxial is used to adjust the quantity of stereo depth in an image (called “Depth Budget” or “Parallax Range”). But note that a lot of people tend to speak of “interocular” (or “I/O”) when they want to speak about interaxial. Some speak about “stereobase” - it's all the same thing: the distance between the center of the lenses of the cameras.

The interaxial varies from shots to shots, and may be changed during a shot:

- When we need a large interaxial (landscape shots, miniaturization effect), we use a side-by-side rig. But this kind of rig do not allow small interaxials, because the bodies of the cameras and/or the lenses collide.

- When we need small and medium interaxials, less than about 14cm/5.5in (which means: for most shots), we use mirror rigs, also called beam splitter rigs. A semi-transparent mirror is set at 45° in front of one camera, which shoots through it, while the other camera is set vertically, on top of the mirror, and shoots in reflection. This system allows very small interaxial distances, as low as 0cm/in if necessary (which is extremely useful to check the proper alignment of the cameras).

The interaxial being useful to set the parallax range properly, the question is: “what kind of interaxial do we want, to get what kind of parallax range?”.

The parallax range (often called “Depth Budget”, please see What's the difference between Parallax Range and Depth Budget? to know why we choose to use “Parallax Range”) is a way to quantify the amount of stereo depth in an image. It is the difference between the parallax of the farthest object and the parallax of the closest object.

For exemple, if in my image the farthest object has a parallax of +1%; and the closest object a parallax of -2%, then the parallax range in that image will be 3% (+1 - -2).

And if the farthest object has a parallax of +2% and the closest a parallax of 0%, then the parallax range is 2% (2-0).

For small screens, for most shot the recommended parallax range varies from 1.5% to 3%. The rules of the 3D channel Sky 3D recommand that the 3D should be most of the time between +2% for far objects and -1% for close objects (that's 3% parallax range). They say that some rare shots with special depth effect could be as deep as 6.5% parallax range: +4% for far objects, -2.5% for close objects - these are very powerful values. For more details, see http://introducingsky3d.sky.com/a/bskyb-3d-tech-spec/ .

For bigger screen, theater screen or IMAX screen, smaller parallax ranges are often recommended: from 0.75% to 1.5%.

Naturally, the exact appropriate parallax range depends on each shot, and depends on the effect the director wants to achieve.

The Closest Object Method - controlling the Parallax Range
The Screen Plane Method

Building the Stereo 3D in an image does not only mean “not making the audience diverge” and “not creating too powerful 3D”…

Stereoscopic and non-stereoscopic depth cues
Aligning the cameras
Synchronizing the cameras
What's the difference between Parallax Range and Depth Budget?
Miniaturization and Giganticism
Focal length
Lighting
Going even further and breaking the rules - the tolerance to divergence

What's different when editing a 3D movie?
Timing
Dynamic Depth Matching
Dynamic Floating Window
The color correction of a S3D film