The document discusses stereoscopic 3D production. It covers the differences between 3D and S3D, natural depth cues, depth perception, the business case for 3D, 3D storytelling techniques, stereoscopic technology formats, live 3D engineering challenges, Sky 3D broadcast models, S3D cinematography theory, S3D pre-production including depth budget and script, stereoscopic editing, and summarizes stereoscopic 3D tools.
1. Stereoscopic 3D Workshop
05.01.01
CRAIG LEES RAVENSBOURNE MA
BROADCAST FUTURES
Stereoscopic 3D production
Difference between 3D and S3D
Stereoscopic 3D is based on the illusion of recreating Human binocular Vision
by using two sets of images captured with slight positional disparities to create
perceived depth. This effect is known as Stereopsis.
3D describes real or computer-generated three-dimensional objects and is also
the term more often used with CGI, 3D modelling and VFX.
Natural Depth Cues
Our brain interprets depth through a number of differing two-dimensional
depth cues such as occlusion. When objects are covered by other objects, our
brain registers them as being nearer/further away to us.
Other natural depth cues include relative size of objects, saturation (objects’
colours desaturate the further away they are).
Depth Perception
Convergence/ Divergence
Parallax – Negative and Positive
The business case for 3D
With the success of Avatar at the Box Office and the Sky Pub 3D experiment, 3D
movies are noticeably attractive as ticket prices are elevated and 3D in the home
naturally has extra revenues for income for manufacturers.
2. Aside from monies from 3D film ticket sales at the cinema, there are a plethora
of 3DTV channels set to hit our screens over the next year.
3D storytelling
S3D has a great opportunity to produce different creative tools. Not only can we
make things appear come out of the screen and into the screen, we can also
make objects appear smaller or greater than what they are (dwarfism and
giantism).
3. Stereoscopic Technology
3D or not 3D?
As of yet there is NO universal standard or format.
Public adoption of 3DTV is an unknown entity. The screens are relatively
expensive with many around the sum of £2000.
Examples of differing formats:
Circular polarising Passive technology
Active Shutter Technology
4. Glasses-free Autostereoscopic Displays using Lenticular or parallax barrier
technology from Toshiba, expected to be released at Christmas
Stereoscopic 3D laptop
Live 3D –
The engineering challenge of getting it right in real-time
Telegenic UK Outside Broadcast provider commissioned by Sky 3D to deliver
3D broadcast using Sony cameras and 3ality systems
Digital technology innovation has made Live 3D possible.
Sony System: SIP, P1 / HDC 1500 camera systems, Passive 3DTV monitoring
7. Sony HD CAM SRW-1 Deck – records both video feeds onto one multiplexed
format
3D Village and Video Village
8. 3ality 3D monitoring on a 2D screen using luma difference 2% parallax grid
3ality 3D monitoring metadata
9. 3ality 3D monitoring alignment and depth budget
Fusion Pace Rig – Engineered by Vince Pace. This system was used on
Avatar, The Final Destination plus recently used on the upcoming TR2N 3D.
James Cameron on Pace Fusion 3D rig
10. Pace On Location
Arri Alexa Cameras mounted on Pace Fusion 3D Segway Rig on Martin Scorcese’s
The Invention of Hugo Cabret at Sheperton Studio
11. The Sky 3D Broadcast Model
Sky 3D Pub Football
12. The Stereoscopic encode format is Side by Side compressed within a 1080i25
frame
To enable the 3D program to retain the highest quality throughout, a minimum
of 90% must be native 3D footage.
Stereoscopic Alignment
These guidelines are for the final Program content being displayed on screen
sizes in the range of 46" to 70":
Main subject point should nominally be the screen focus point or
convergence point of the two images
Positive disparity or image separation at distant points (into the screen)
should not exceed 2% for majority of shots
Negative disparity Image separation at close points (Out of Screen) should not
exceed 1% for shots.
Record format should be at a bit rate no less than 100Mbit/s 4:2:2 for acquisition.
S3D Cinematography Theory
The Stereo Window – think about shooting within a window to avoid edge
violations. Monitor BOTH cameras.
Calculating 3D depth for a scene – the variables of screen size
Converged cameras Vs Parallel cameras – weigh up the pros and cons for what
is best for your shoot
Pre-visualisation – weigh up the best solution for your production – DOES NOT
HAVE TO BE EXPENSIVE!
Controlling depth – The Depth Bracket and depth script. How much depth do
you want in the scene to express your storytelling? What should be the Inter
axial distance: Hyper stereo, Ortho-Stereo or Hypo stereo? Pre-production is
VERY important in Stereo 3D.
The effect of screen size upon the 3D effect – What is your final output? This
impacts on whether you should employ 1/30th / 1/100th laws for stereoscopic
cinematography.
13. The importance of Metadata – When using VFX, it is very important that you
record stereoscopic camera data for your VFX artist to match effects. What is the
IA? What is the distance to nearest/furthest distances? What is the focal length?
What is the F-stop?
Twice the number of cameras is required for a native 3D stereoscopic 3D camera
system. This results in TWICE THE DATA.
Zoom Lens problems: telecentricity errors, also known as zoom wander.
The effect of different focal lengths. Wider lenses create a greater 3D effect
whereas longer lenses tend to flatten the 3D effect.
Lighting. Reflections, high contrast imagery shadows and specular highlights
often create visual artefacts in stereo 3D
The importance of using shutter-sync cameras
Cameras with Genlock sync
Figure 1, R Cam 1 frame before clap L Cam 1 frame before clap
Figure 2, R Cam at clap L Cam at clap
14. Cameras with Temporal sync (shutters NOT synced)
Figure 3 L Cam 1 frame before clap R Cam 1 frame before the clap
Figure 3, L Cam at noise of clap R Cam at noise of clap
Fast moving objects in temporally synced cameras can cause Retinal Rivalry
artefacts, as we are seeing different images between our left eye and our right
eye, which can cause discomfort. The greater the size of the screen, the more
retinal rivalry is amplified.
S3D Pre-Production
Depth Budget and Depth Script
Each scene will have a designated depth budget associated with it for a
stereographer to work with on set. This budget will be created using a variety of
factors including feel, continuity, and broadcast/cinema specifications. A depth
script will be constructed in the post-production process and are likely to have
decisions regarding Inter-axial and convergence settings. Note that though
errors in convergence can be easily corrected in post, altering the inter-axial
distance is very difficult. To date, only highly expensive systems such as Mistika
15. and Occula are capable of changing inter-axial distance in a scene, furthermore,
the large computational data required by this interaxial operation means that
these advanced programs cannot render these actions in real time.
The depth script is constructed in the form of a graph that tracks the filmmakers
emotional intended impact of each scene. These depth decisions can be created
16. in the script with signifying notations such as; C-5 (lenses strongly converged
so that the action very much happens in the viewer’s lap – lots of negative
parallax); C-0 (placing key subjects on the screen plane); and C+5 with
moderately converged lenses pulling the viewer into the screen space.
As well as convergence decisions, another major component of the depth
budget is the required interaxial distances that each scene should contain,
which again is very much part of the pre-production process. Due to inter-axial
distance being the single major influence over perceived 3D depth, attention
should be noted upon interaxial distances and that they do not jump from one
extreme to the other. Such extremes produce uncomfortable depth cuts upon
the viewer. Also, interaxial distances have an effect upon scale. This sense of
scale, due to exaggerated inter-axial distances, brings filmmakers the
opportunity to enhance the emotional impact of their stories (Alice in
Wonderland, where Alice grows taller and shrinks being a great example).
Narrow inter-axial distances invoke gigantism effects (making the subject
appear larger than what they are). Wider interaxial distances cause
miniaturisation effects (making subjects appear smaller than what they are).
These scale decisions can be notated in the script much in the same way
convergence notations do such as I-5; I-0; I+5 , where I-0 indicates orthostereo
(human interpupillary distance – 65mm or thereabouts), I-5 indicates the
interaxial where the lenses are narrowly separated (hypostereo) and I+5 where
the lenses are widely separated (hyperstereo).
The depth budget follows the decisions made within the depth script and is
maintained by live 3d pre-visualisation measuring tools such as the 3ality
Stereo Image Processor (SIP 2100) and the Sony MPE-200 SIP, or by simply using
a grid overlay on the multiplexed stereoscopic imagery.
17. Things to consider
3D uses double the bandwidth of usual video editing as two video tracks are
used to create every S3D shot. Spec your machine for a faster stereoscopic
pipeline.
Depth Cuts
Floating Windows.
Slower cuts for shots with lots of perceived 3D depth
Multiple S3D viewing and output modes are very useful
Organise your media.
Products
Cineform (Mac & Windows) Neo 3D around £2000 and Neo HD (under £400)
Stereo 3D Toolbox (Mac) Final Cut Pro Plugin created by Tim Dashwood (around
£800)
Avid Nitris around £10K
Mistika around £100K
Quantel Pablo 3D around £100K
Smoke (Mac & Linux) around £10K
After Effects CS5 scripting and expressions
Premiere CS5 – now has stereo functions and can handle 4K Red footage (with
Nvidia Quadro Graphics Card) – VERY EXCITING! Shows how powerful non-linear
edit systems (NLE’s) are becoming.
DIY 3D - Stereoscopic 3D on a budget
Build your own rig
18. There are benefits in shooting parallel
Stereoscopic player – available from 3tv.at Dual firewire
Neo Player – Cineform’s Player allows you to play the Cineform S3D Codecs on
your machine using a 3D display.
Adapt your existing HD kit
Vizard – Build your own 3D theatre – Wired magazine article
Zalman Monitor – Full colour Passive S3D under £300
Anaglyph monitoring – 3D monitoring on a 2D screen (VERY cost effective)
NVidia 3D vision solution
iZ3D – stereoscopic 3D gaming on a PC
Stereo 3D Toolbox by Tim Dashwood
Today we will learn the basics of the FCP plugin
Stereo 3D Toolbox.
Why? It is VERY stable and has lots of fine
extensive stereo tools, which can produce
everything you need for your S3D productions.
Exercise One
Making a Stereopair
Sync your clips together for your Left Eye Clip and
Right Eye Clip. Make sure they have the same
start and finish points. If they don’t, trim them.
19. Once you have ensured they are temporally
synced and are the same duration, double click
each clip and then Make a Subclip (Command U
for you hot key freaks out there – like me!).
This will make a subclip in your browser.
Do the same for your Right Eye temporally synced
clip.
NOTE – Why do we have to make subclips?:
Well, FCP recognises start media and end media
only in your video clips. It is very difficult to get
two cameras two record at EXACTLY the same
moment of time. Making subclips is the fail-safe
way of ensuring your clips start and end at the
same point in time. Setting In and Out points alone
does not work!
Once you have made a Left and Right Subclip,
which you can see in your browser, we can now
make our stereopair.
20. NOTE – your clips will be named depending on
how your media is captured. When using
tapeless, make sure you use Log and transfer to
name your clips - TRUST ME - this will save you
lots of pain later!!
Then, drag your Left eye clip into your timeline.
Once there, double-click your Left eye clip, then
go to Effects – Video Filters – Dashwood Stereo
3D Toolbox (NOT the LE lite edition) – Stereo 3D
clip coupler.
21. Click the Filter Tab. On the S3D Input, Select Clip=
Left eye & Image Well = Right Eye.
22. This will enable your Left Eye Clip to be the
Master Clip and it will create an Image Well that
you can drag and drop your Right Eye Subclip
into to create your stereopair. Keep the S3D
Output Mode as Side by Side at this stage. If you
have your Left and Right Eyes the wrong way
round, You can select the Swap Output tab, but
getting it right first time is VERY GOOD PRACTICE!
23. Once you have created your stereopair using Clip
Coupler, you then want to geometrically align
your clips in order to produce the desired
stereoscopic effect.
To do this we double click our clip – Effects –
Video Filters – Stereo 3D Toolbox – Stereo 3D
Geometry.
24. Click on the filters tab for the stereopair clip, if not
already, to reveal the Stereo 3D Geometry
Controls.
25.
26. Once the Stereo 3D Geometry Controls Tab is up,
we can move the images around to create the
aspired stereoscopic effect. This is the fun part as
this is where the S3D comes alive.
Depending on how you have shot your
stereoscopic material, whether it be parallel or
converged, the amount of applied geometrical
alignment will change. If the material has been
shot using the parallel method, we will need to
apply what is called Horizontal Image
Translation (HIT) to the two images, which
basically involves sliding the two images
27. horizontally over each other until they slightly
overlap to the point of convergence that best
suits the material.
Anaglyph Stereo pair with no HIT applied
Anaglyph Stereo pair with HIT applied
28. To adjust HIT, use the Convergence Slider in the
Stereo 3D Geometry panel and change the S3D
Output mode to the appropriate 3D viewing mode
you are using, this may be interlaced for passive
screens such as the Zalman, side by side for 3DTV’s
such as the JVC or anaglyph if you are using a
standard 2D screen.
Use the Convergence % Course slider first to get as
close to the required convergence point possible,
then use Convergence % Fine to make finely
tuned, more accurate alignment adjustments.
If you have vertical misalignments, use twirl down
the Disparity Correction triangle to release the
vertical alignment controls.
29. Other misalignment errors that you may wish to fix
include Rotation, Zoom, Keystone (these are all
quite common), can all be fixed here in the Stereo
3D Geometry controls.
30. How do I measure my 3D effect?
There is a great tool in the Video Generator tab
which can be accessed by selecting - Dashwood
Stereo 3D Toolbox – Stereo 3D Parallax guides.
31. These Parallax Guides are REALLY useful as a
means to measure how much depth you have
coming OUT of the screen (Negative Parallax) and
back into the screen (Positive Parallax) within each
scene.
And this is where good old anaglyph monitoring is
really useful.
TIP – Negative Parallax (coming out of the screen),
the disparity ‘fringing’ will be GREEN on the left.
32. Positive Parallax (going back into the screen), the
disparity fringing will be RED on the left.
If it’s not, you have got your eyes the WRONG WAY
ROUND! Don’t stress though, this is a common
mistake when first starting out!
You can set the guides to a matter of percentage,
which is good, but make sure you are monitoring on
a screen the same size of your final output screen.
The Sky 3D Guidelines are 1% negative parallax and
2% positive parallax, which is a pretty good rule to
stick to though you may want to push this further
when displaying on smaller screens below 50”.
33. To select parallax guide controls, click the controls
tab, then select custom parallax %, then change the
slider (or type in the value) to 1% custom parallax.
34. Then drag this video generator onto your timeline
above your stereopair.
Now you can measure your 3D! Measure the fringing
disparity of the object closest to the camera to
measure your maximum negative parallax (furthest
out of the screen object which has green fringing on
the left), and measure your maximum positive
parallax (the furthest behind the screen object
which has red fringing on the left).
What is the measured 3D Depth in this scene?
35. Yes you can change the colour of the parallax guides – Very
useful!
Expanded view of the open window. The yellow marks show how the
stereo infinity positive parallax divergence rests measures 1%.
36. Analysis of Scene
Ma ximum p ositiv e p a ra lla x/d isp a rity 0.8 %
Ma ximum ne ga tiv e p a ra lla x/d isp a rity 1%,
Based on: IAD 20mm, focal length 7.6mm, convergence point actress
(2m), closest point shaven-headed actor's shoulder (1.6m) , furthest point
tree-line above house outside (50m)
……………………………………………………………………………………..........
Ok, that is the end of this introduction to the Stereo3D Toolbox
plugin and Stereoscopic fundamentals.
There’s more than enough here to keep you going for some time in
your beginning adventures with S3D. There are plenty of other useful
tools within the plugin such as creating 2D text in 3D space and
Displacement Maps but for now concentrate on these tools to build
your stereoscopic stories.
And questions or help required, email me at cogartlit@yahoo.co.uk
and I will try to answer your questions as quickly as possible.
Thanks and good luck new 3D movie-makers!
Craig Lees
Ravensbourne MA Broadcast Futures
Further research material
Bernard Mendiburu’s 3D Movie Making from Script
to Screen.