Stereo 3D Filmmaking (http://www.convergence3d.net/stereo3d.php)

Has anyone watched this and highly recommends it?

I already have Stereoscopic Digital Cinema from Script to Screen
by Bernard Mendiburu and I was curious to know if there was anything new or useful this interactive course may teach.

The $400 price point is pretty steep for only 2-hours of training plus 6-hours of expert 'insight', can't really tell how useful that may be.

best,

I haven't seen it and it doesn't say who the instructor is but I would imagine that the "expert insight" DVDs are worth more than going to a weekend course that would probably set you back more than $400.

Bernard's book is awesome and mostly up-to-date as of last Fall, but it won't teach you "how" to shoot stereoscopic 3D. It covers everything else except the math required to choose the proper lenses and interaxial distances. You should download the pdf of Lenny Lipton's Foundations of Stereoscopic Cinema and read chapter 3 to learn about the various equations you can use to keep your maximum positive parallax in check. Be sure to reference the errata list at the end of the book because there are quite a few errors in the algebraic equations.

Hey Tim,

I've actually read all of lenny lipton's books on stereoscopy. Pretty interesting stuff, a lot of it was pretty damn complicated with all the math equations. My math goes as far as 2 + 2 = 1. But maybe i'll give it another read since i read it a year or two ago.

I was also thinking that the 6-hours of 'expert insight' would be pretty worth it, but hard to tell. I wish they had some samples they could share.

If i end up getting it, i'll let you guys know how it turned out.

Stereo 3D Filmmaking (http://www.convergence3d.net/stereo3d.php)

Has anyone watched this and highly recommends it?

I already have Stereoscopic Digital Cinema from Script to Screen
by Bernard Mendiburu and I was curious to know if there was anything new or useful this interactive course may teach.

The $400 price point is pretty steep for only 2-hours of training plus 6-hours of expert 'insight', can't really tell how useful that may be.

best,
I find it totally useless as both BM and LL have no basic understanding of human stereoscopic vision.
You will not find a book or tutorial that describes stereoscopic movie making that is screen size independent and yields full realistic immersion with out geometrical distortions.

Mathew Orman

You will not find a book or tutorial that describes stereoscopic movie making that is screen size independent and yields full realistic immersion with out geometrical distortions.

Maybe that is because stereoscopy is screen size dependent? If you shoot for a 10-foot screen and project it on a 20-foot screen, your interaxial distance becomes too large, so your audience is somehow supposed to twist the left eye to the left and the right eye to the right, something we are not biologically equipped to handle.

And if you put it on a DVD or BD, the interaxial distance becomes too small when viewed on a 32" screen. Worse yet, some people will watch it on a 40" screen, and others on a 22" screen, etc. At least our biology allows us to watch that but it does not look the same, stereoscopically speaking, as the 10-foot screen.

A possible theoretical solution would be shooting extra width, which would require specialized cameras, such as the Red One, so, for example, when shooting at 1080p your width would be more than the standard 1920 columns. Let's say your acquired width is 2000 columns, so you have 80 extra columns of width. In post you choose the right 1920 columns for the big screen and a different 1920 columns for the TV or computer monitor.

In other words, you bring your left and right views closer for the big screen, closer yet for the very big screen, farther for the small screen.

Maybe that is because stereoscopy is screen size dependent? If you shoot for a 10-foot screen and project it on a 20-foot screen, your interaxial distance becomes too large, so your audience is somehow supposed to twist the left eye to the left and the right eye to the right, something we are not biologically equipped to handle.

And if you put it on a DVD or BD, the interaxial distance becomes too small when viewed on a 32" screen. Worse yet, some people will watch it on a 40" screen, and others on a 22" screen, etc. At least our biology allows us to watch that but it does not look the same, stereoscopically speaking, as the 10-foot screen.

A possible theoretical solution would be shooting extra width, which would require specialized cameras, such as the Red One, so, for example, when shooting at 1080p your width would be more than the standard 1920 columns. Let's say your acquired width is 2000 columns, so you have 80 extra columns of width. In post you choose the right 1920 columns for the big screen and a different 1920 columns for the TV or computer monitor.

In other words, you bring your left and right views closer for the big screen, closer yet for the very big screen, farther for the small screen.

I have been trying to understand this concept and don't think I quite have it, but it would seem to me that it is not necessary to produce a 1920 wide shot, an 1840 wide shot would be fine and you could do what you are saying with conventional HD cameras. I also am wondering if this is something that can be done, pixel selection, at the time of playing back the video such as in a Blu-ray player or PC. I need to reread your Seven masters link but I would think this might be something encoded in the file data, screen size, so that the player would know what pixel offset to use. As I said I am not really clear on the issue and seem to be a slow learner in this area. Thanks for your insight on screen size affecting the video display Adam.

I would think this might be something encoded in the file data, screen size, so that the player would know what pixel offset to use.

Indeed it should. I do not mention it on the Seven Masters site (I should have but was not thinking about it when I was finishing the page). I intend to include this detail in the Seven Masters API, however.

Though, most video file formats, at least for Windows, have a way to include a form of "DPI" which defines the size the images are designed for when displayed. Then again, most software pretty much ignores this value.

Here is how the problem works: Suppose you use the standard 1920x1080 frames and you edit your video on your computer with a screen that is 38 cm wide (just using numbers that make the math easy), so you get roughly 50 columns per centimeter. So, if you make the largest distance between an object in the left and right views 60 mm (6 cm), you can go as far as 300 columns apart, in theory, and you can handle viewing it. But watch it on a 19 m movie screen, with about one column per centimeter, and now the difference is 300 cm, which is completely impossible to view in 3D.

Indeed it should. I do not mention it on the Seven Masters site (I should have but was not thinking about it when I was finishing the page). I intend to include this detail in the Seven Masters API, however.

Though, most video file formats, at least for Windows, have a way to include a form of "DPI" which defines the size the images are designed for when displayed. Then again, most software pretty much ignores this value.

Here is how the problem works: Suppose you use the standard 1920x1080 frames and you edit your video on your computer with a screen that is 38 cm wide (just using numbers that make the math easy), so you get roughly 50 columns per centimeter. So, if you make the largest distance between an object in the left and right views 60 mm (6 cm), you can go as far as 300 columns apart, in theory, and you can handle viewing it. But watch it on a 19 m movie screen, with about one column per centimeter, and now the difference is 300 cm, which is completely impossible to view in 3D.

I understand that portion but if you are sitting back from the screen an appropriate distance wouldn't that large separation, 300 cm, be correct for that particular object's location/depth? I don't want to hijack this thread for this conversation and will hopefully gain some understanding from Lipton's book that will make this more clear to me, so no need to respond to my question. Thanks for your response and the original question of image produced for a particular screen size will be a better topic in your seven masters thread.

if you are sitting back from the screen an appropriate distance wouldn't that large separation, 300 cm, be correct for that particular object's location/depth?

No, it would not. No matter how far you sit, your eyes still have to diverge. The angle of divergence will be smaller at the large distance, but it is still a divergence. And our eyes are not built for divergence, only for straight (parallel) and convergent viewing.

In real life, we never have to diverge our eyes, no matter how far the object we are looking at is. When an object is close, we converge the eyes (bring them closer, like when you put a finger near the tip of your nose), when it is far, our eyes converge much less (they can even be parallel). But they never ever diverge (look in the opposite direction).

So, to view the stereo on a large screen with the objects too far you essentially have to wiggle your eyes at all times, looking at one view of the object one moment, the other view the next moment, and that is the main cause of the eye fatigue many people report after watching a 3D movie.

Hi

The screen size problem is indeed very important in Stereo 3D.

Adam, you're right, one can modify the HIT (Horizontal Image Translation) or "Post-production Shift" to accomodate a 3D movie for either a small or big screen. The HIT simply shifts left and right images with respect to each other.

In theory (although it isn't always done), one should have a "small screens version" (computer screens/TVs) and a "large screens version" (theaters). Sometimes there are more versions (ex: very larges screens, imax version)

To answer more precisely, here is the terminology that seems to be becoming a standard:

1) an object's "parallax" is the distance, in % or pixels, between its left and right images (usually between -2% to 1%, sometimes - very rarely - as big -4% to 2.5%). Negative parallax is "crossed parallax" (left image on the right, right image on the left) and appears in front of the screen, positive parallax is "uncrossed parallax" and appears behind the screen.
2) an object's "screen parallax" is its parallax "on the final projection screen", measure in centimeters or inches
3) the "total parallax range" of an image is a measure of its total depth: it's the difference between the parallax of the farthest object, minus the parallax of the closest object
4) "interaxial" is the distance between the cameras when shooting

Notes:
a) Parallax range is dependent on interaxial and is very difficult to change in post-production, so it remains the same between small and large screen versions.

b) Some recommend parallax ranges between 0.5%-1.5%, others say that's "flat 3D" and recommend 1.5%-3%. Sky Broadcasting recommends 3% (and they even talk about 6.5% for very specific shots, although that might be risky). Usually, 3% works for small screens but might be difficult to watch for bigger screens. 1.5% might be better for theaters. (with the right H.I.T., of course)
All this is partly a question of taste, and depends on types of shots, its non-stereoscopic cues like geometry, lines, depth information etc. and also on previous shot's depth matching

b) Although parallax range doesn't change with versions, the "far parallax" and the "near parallax" shift with the chosen H.I.T./Post-production Shift.
Example: with 2% parallax range, smaller screens might prefer 1.5% far parallax, 0.5% near parallax (out-of-screen effects), but on theater screens wider than 10 meters this will diverge so large screen version would put 0.5% parallax (or less) and 1.5% out-of-screen.

c) most stereographers strongly recommend avoiding screen parallax (or screen offset) going over 6.5cm/2.5 inches, the average distance between human eyes. There is some debate around this issue (cf. "waking up hangar" in Avatar). Most people accomodate some limited degree of divergence (0.5 to 1.5 degrees), although depends on people and on viewer distance to the screen. Some use that, some consider it risky in terms of eye strain.

Here are SKY's recommendations: (they have slightly different vocabulary: "Disparity" is the more scientific term for "Parallax")

------------------- SKY GUIDELINES
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 be used with care and not nominally exceed 1% for shots. Care should be taken for images breaking the frame edges with floating windows utilized where appropriate
These are guidelines that aim to deliver managed and comfortable stereoscopic viewing. As such these limits can be exceeded for specific editorial needs, (Such as Graphic Content or Short Term visual impact), managed appropriately and in line with 3D production practice.

Such instances should be constrained to 4% Positive (Into Screen) and 2.5% Negative (Out of Screen)
-------------------
source:
BSkyB 3D technical specification | Sky 3D - a new dimenstion in Sky+HD | Sky (http://introducingsky3d.sky.com/a/bskyb-3d-tech-spec/)

I hope this clarifies a bit.

One last thing:

I'm the writer and co-director of the course discussed in the title of this thread: "Stereo 3D Filmmaking: The Complete Interactive Course".

To answer the first post, I honestly think it's actually quite inexpensive compared to the amount of material that's covered, and compared to training workshops. Seriously :)

We try to provide an in-depth view on stereo-3d shooting and principles of post-production that really cover a broad range of techniques so one can choose based on personal style or type of S3D shooting. It's very in-depth, it took us a year to post-produce, we put a lot of attention into every detail.
It goes from basic principles to very advanced techniques and is very, very dense. There is 6-8h of exercises in a virtual studio based on award-winning Frameforge Previz Studio (its makers Innoventive Software are co-producers of it with our company Parallell Cinema) and the Expert Insight Series come from real people in the business in their own field of expertise (Sony Imageworks' VP for 3D Production, 2D-3D conversion from the people who did Alice etc.)

If you want more info:
- Convergence 3D (http://www.convergence3d.net/) for a free knowledge base/discussion area/job listings etc.
- Stereo 3D Filmmaking: The Complete Interactive Course (http://www.convergence3d.net/stereo3d.php) for more info about the "Stereo 3D Filmmaking" course itself, a trailer and worldwide ordering information.

That's it!

We sincerely hope both Convergence3D.net and our course will help filmmakers switch to 3D in a both easy and professional way.

David Steiner
www.parallellcinema.com

Thanks, David, and welcome to the forum.

Hi Bicky,
I recently purchased this DVD course and just finished watching it. My reply may not be entirely useful, as I *DO* recommend it, but not "highly," as you asked. My main criticism is the price.

The two hours of lecture is crammed full of information and examples! I liked seeing immediate examples of what is being taught. The 3D is clean and inspiring.

On the negative side, the lecture is really just two large movie files. I wish they would have split it up into lessons with a better interface, and then maybe had an interactive Q&A session after each lesson, where the user must choose the correct answer. The host's accent was a bit distracting to me at first, too. Two of the four DVD's were just interviews, which I personally did not find useful at all.

One of the most-unique features is the "FrameForge Stereo Lab," where they have created some virtual sets, and you can move the cameras around and see the stereo result immediately. There is an audio guide to walk you through the examples, which is quite nice. But for some reason, I thought I was getting an actual copy of FrameForge. As far as I can tell, there is no way to create your own set without buying a FrameForge upgrade, so I grew tired of the lab once I had learned the concepts.

As David points out, it is priced below any other "comparative" course. But as a self-taught student of stereography (meaning I do it as a hobby and don't get paid) I can't quite justify the price. If the course were priced at $200 or less, I'd tell everyone I know to go buy a copy!

Bottom line: I recommend it, but I do think it is over-priced (for me). But if you are getting paid for your stereo 3D work, it will pay for itself very quickly, I'm sure.

I run workshops in the UK and charge £150 GBP (about $250USD). You get a full day course combining theory sessions and practical hands on sessions. I've also run courses in Singapore and Switzerland. In July I'll be doing courses in Prague, August in South Africa and the October India, Indonesia, Singapore, Malaysia, Thailand and Hong Kong.

Still the best way to really understand stereoscopic production is to get a pair of cameras and experiment. A couple of Go-Pro's and a simple mounting bar would allow you to experiment with camera separation, convergence and post production. IMHO you can't beat doing it for yourself, making mistakes and learning from those mistakes. Stereoscopic production is difficult to get your head around as many aspects are counter intuitive. Going out and doing a shot then figuring out why it does or does not work is a great mental exercise. Books, DVD's and courses are useful but there's no substitute to getting out there and shooting some clips for yourself.

Hey all, I'm the designer of the FrameForge Previz Studio Lab with the course and my company, Innoventive Software, was one of the co-production entities on the film, along with Parallell Cinéma.

Sorry for any confusion about the Laboratory -- it is designed to be just that: a laboratory where you can explore and experiment with the concepts taught in the course, and take it from a pure lecture into a real, hands-on experience. But as you said, it's not designed for you to create your own new scenes.

We've tried to make it clear that it wasn't the full the Stereo 3D version of FrameForge Previz Studio, which itself sells for $899, but if you have any suggestions as to what confused you, we'd like to know.

As to the price, as you alluded, it IS a professional training class and we believe that for the quality, quantity and depth of information plus the unlimited hands-on access to a (virtual) stereo 3D rig makes it well worth the money.

Maybe that is because stereoscopy is screen size dependent? If you shoot for a 10-foot screen and project it on a 20-foot screen, your interaxial distance becomes too large, so your audience is somehow supposed to twist the left eye to the left and the right eye to the right, something we are not biologically equipped to handle.

And if you put it on a DVD or BD, the interaxial distance becomes too small when viewed on a 32" screen. Worse yet, some people will watch it on a 40" screen, and others on a 22" screen, etc. At least our biology allows us to watch that but it does not look the same, stereoscopically speaking, as the 10-foot screen.

A possible theoretical solution would be shooting extra width, which would require specialized cameras, such as the Red One, so, for example, when shooting at 1080p your width would be more than the standard 1920 columns. Let's say your acquired width is 2000 columns, so you have 80 extra columns of width. In post you choose the right 1920 columns for the big screen and a different 1920 columns for the TV or computer monitor.

In other words, you bring your left and right views closer for the big screen, closer yet for the very big screen, farther for the small screen.

Here it is how human stereoscopic vision works:
65 mmm base stereo camera captures a scene with normal 35 to 55 horizontal view angle then the projection systems projects the scene on a screen that has arbitrary size and the same matching horizontal view angle
the scene's stereo window is always at it's physical position regardless of the screen distance and or size.
Correct projection geometry is the key to realistic immersion stereoscopic experience and it is the same from IMAX to cell-phone screen sizes.

The problem again is the lack of knowledge of correct stereoscopic vision geometry which results in production of totally distorted stereoscopic content projections which it turns proves that current S3D is a nothing but a gimmick that increases the profits on initial runs.

To persist on the long run S3D technology must include undistorted content projection and 2D compatibility.

Mathew Orman

I understand that portion but if you are sitting back from the screen an appropriate distance wouldn't that large separation, 300 cm, be correct for that particular object's location/depth? I don't want to hijack this thread for this conversation and will hopefully gain some understanding from Lipton's book that will make this more clear to me, so no need to respond to my question. Thanks for your response and the original question of image produced for a particular screen size will be a better topic in your seven masters thread.

Yes,
you are getting closer to full understanding of human stereoscopic vision geometry.
The 65 mm parallax is a false assumption as human eyes are perspective cameras and not orthographic and the 65 mm vanishes to single point in perspective geometry. Correct stereoscopic projection geometry includes
vanishing point geometry and assures that the captured scene is observed at the same physical size and distance as the stereoscopic camera that had captured such scene regardless of screen size.

Mathew Orman

Your first false assumption is that the rules that apply to viewing the actual 3D space apply to the projection of two views on a flat 2D screen. The rules are different for both.

The second false assumption is that the distance of the viewer from the screen is proportionate with the width of the screen. This is not true if for no other reason that not everyone sits in the back row, especially not in a crowded theater. Depending on the size of the theater, the persons sitting in the front row sit anywhere between 20 (quite common) to several hundred (rare but I have seen such theaters) times closer to the screen than the persons sitting way back. Additionally, the persons sitting at the side of the front row have a completely different view from the one person sitting in the center of the front row.

Your eyes must never diverge, we are just not biologically and physically suited for that.

You have to adjust the parallax based on the width of the screen. Either that or you somehow have to sculpt a real 3D image in actual 3D space, something we certainly do not know how to do. Maybe someday we will. But as long as we are projecting two separate views on a 2D screen, 3D video is dependent on the size (well, width) of the screen.

Your first false assumption is that the rules that apply to viewing the actual 3D space apply to the projection of two views on a flat 2D screen. The rules are different for both.

The second false assumption is that the distance of the viewer from the screen is proportionate with the width of the screen. This is not true if for no other reason that not everyone sits in the back row, especially not in a crowded theater. Depending on the size of the theater, the persons sitting in the front row sit anywhere between 20 (quite common) to several hundred (rare but I have seen such theaters) times closer to the screen than the persons sitting way back. Additionally, the persons sitting at the side of the front row have a completely different view from the one person sitting in the center of the front row.

Your eyes must never diverge, we are just not biologically and physically suited for that.

You have to adjust the parallax based on the width of the screen. Either that or you somehow have to sculpt a real 3D image in actual 3D space, something we certainly do not know how to do. Maybe someday we will. But as long as we are projecting two separate views on a 2D screen, 3D video is dependent on the size (well, width) of the screen. """"

I do not have to do anything.
The are no assumptions. It is strict geometry of viewing system.
In theaters there is only one seat that matches the view angle and stereo-window distance.

Mathew Orman

In theaters there is only one seat that matches the view angle and stereo-window distance.

And do you want to shoot 3D movies that you show to one person per theater? Maybe you do, maybe you do not.

But if you shoot movies to show to full theaters, as most filmmakers do, you need to deal with the parallax.

No,
in my theaters viewers have glasses that match their position geometry with projection geometry.
Most directors still shoot 2D with distorted perspective and that fallacy
continues into S3D shooting resulting to total gimmicks and zero realistic immersion in original scenes.

Mathew Orman

No,
in my theaters viewers have glasses that match their position geometry with projection geometry.

And what percentage of theaters do you own?

This is not about you. This is about reality.

And what percentage of theaters do you own?

This is not about you. This is about reality.

None,
I was referring to my new S3D vision system for which I've just filed a US patent application.
System which is at least 400% brighter , zero color distortion, zero ghosting, zero flicker and full realistic immersion on any size of a screen and 2D/S3D viewable simultaneously with equal quality regardless of seat position.

Mathew Orman

I understand that portion but if you are sitting back from the screen an appropriate distance wouldn't that large separation, 300 cm, be correct for that particular object's location/depth?

There seems to be some confusion here, which stems from looking at the interaxial distance the way we perceive farther objects as smaller than nearer objects of the same size. This, however, is a completely separate issue.

A distant object appears smaller because the angle needed to see it is smaller than the angle needed to see the same object at a close distance. This is an optical illusion which makes the rails of a railroad connect at a certain distance. It has nothing to do with stereoscopy because the illusion also exists when you close one eye, or when you take a photograph with a single-lens (2D) camera.

In reality, those two rails never connect, they only appear to. You can ride the longest railroad in the world, which, as far as I know, is the Transsiberian Railway. You can board a train in Moscow and travel thousands of miles to Vladivostok, and the rails will keep a constant distance. This is the same as looking for a pot of gold at the end of the rainbow: There is no pot because there is no end of the rainbow. And two parallel lines never intersect even though they appear to.

So, when the two images of the same object are 65 mm apart, we look at them in parallel regardless of what the size of the screen is or how far the screen is. The distance will appear smaller but it is not smaller.

The easiest way to think about it is that two parallel lines appear to come closer and even intersect when we are looking at them, but they stay parallel when we are looking with them or through them. In other words, they appear to intersect when they are the object of observation but they stay parallel when they are the vehicle of observation.

I hope this makes it easier to see the difference and makes it clear that the 300 cm separation is never correct.