3D Stereoscopic Film with HVX200 ?

[dcmbeats]

hi guys,

I'm new in this, so please excuse my ignorance.

I wonder if it is possible to sync two panasonic HVX200 cams, in order to produce a stereoscopic film ?

(I only know the possibility to sync timecode and usersettings with firewire or SD-Card, but this isnt what I need - I think)

Is it right, that it is decisive, that the recording starts exactly at the same time ? (apart from angle and so on...)

thanks for any information

greets jonny

 

[istvanttt]

In the worst case you can use in post the sound-track of the left and right clip to sync. If you don't have fast camera or talent movements then you can stay with it.

hi guys,

I'm new in this, so please excuse my ignorance.

I wonder if it is possible to sync two panasonic HVX200 cams, in order to produce a stereoscopic film ?

(I only know the possibility to sync timecode and usersettings with firewire or SD-Card, but this isnt what I need - I think)

Is it right, that it is decisive, that the recording starts exactly at the same time ? (apart from angle and so on...)

thanks for any information

greets jonny

 

[dcmbeats]

thank you very much for you answer.

you say it: "worst case" ! - is there any - let me say - professional solution, to control both cameras simultaneously ?

 

[Eric Lange]

hi guys,

I'm new in this, so please excuse my ignorance.

I wonder if it is possible to sync two panasonic HVX200 cams, in order to produce a stereoscopic film ?

(I only know the possibility to sync timecode and usersettings with firewire or SD-Card, but this isnt what I need - I think)

Is it right, that it is decisive, that the recording starts exactly at the same time ? (apart from angle and so on...)

thanks for any information

greets jonny

check out www.21stcentury3d.com/technology.html#top

these guys (21st century 3d ) have a nice rig built aroun the AG-DVX100a s.

We develop a new type of sofware that can extract clean 3d geometry and textures from stereo dynamic and stereo stills. Would you have any use of such a software? The software and interface are fully immersive.

We are trying to design systems around the ONE, but we also work with with high res stills cameras. Our data sets are also very VR friendly.

Cheers,

Eric

P.M. about the prices from 21st century 3d (if you are interested).

 

[Benni Diez]

Eric, is there some beta testing? Sounds very interesting.

 

[Eric Lange]

Eric, is there some beta testing? Sounds very interesting.

I will PM you and be in contact, nice show reel.

cheers,

Eric

(I see now that there appears to be a PM restriction on your "account", so will phone/contact you perhaps. My German is little rusty for technical conversations but either way I can e-mail you via more conventional channels.)

 

[Dan Hudgins]

Crystal sync

Crystal sync

hi guys,
I wonder if it is possible to sync two panasonic HVX200 cams, in order to produce a stereoscopic film ?

The crystal sync should hold for 30 seconds to a minute before it gets out of sync by one frame, for 3D work you would want less than one frame sync error since you get 3D artifacts if the cameras are not in sync on fast moving subjects.

All you have to do is use a slate clap board to find the same start frame on all four parts, 2 images, 2 sound.

When you set up to shoot do not turn the cameras off parallel so much that anything in the background will diverge the viewers eyes more than 6.25cm (~2.5 inches) on the background, since if the eyes of the viewer need to diverge too much he will see double or get eye strain. Its better to have the most distant object in the shot line up in both cameras to avoid that at all viewing distances and projection sizes, or just keep the cameras parallel with the normal offset for the focal length used.

If you are outdoors you can line up of something in the far distance, if you are indoors you can tip in a little but keep things on the far walls together. If you tip in on the subject too much outdoors the things at INF may go too far apart on the screen to fuse for most viewers.

 

[Benni Diez]

Thanks Eric.


You will need cameras with Genlock, like the EX3. If you sync the material later you will have a slight sub-frame offset most of the time which will be uncomfortable with fast motion. And even if you sync the record start perfectly, you will get offset after a while. Every frame needs to be perfectly in sync.

I wonder how the semi-pro rigs handle focus.


The P+S rig looks pretty cool btw:

http://www.reel-show.tv/index.html?vidId=00063

Side-by-side rigs are cheap but have very limited control over depth and convergence.

 

[Eric Lange]

Thanks Eric.




I wonder how the semi-pro rigs handle focus.


The P+S rig looks pretty cool btw:

http://www.reel-show.tv/index.html?vidId=00063

Side-by-side rigs are cheap but have very limited control over depth and convergence.

The rig we are building is more of a surveying grade instrument. For focus, the principal distance of the lenses we use is calibrated to +/- 2.5 micron, and the angular precision of the cameras is generally of the order of 7 arc seconds, (very high res encoders), and further possible to get down to 3 arc seconds. For dynamic focus pull we are having to build lens “cones” that have cross roller bearing stages that are actuated by precision ball screws and servo motors, linear encoders and motion controller to effect the same 2.5 micron precision. Special setting techniques are required and from time to time using autocollimators to very accurately set the “start point” of the lenses. We mainly work with Schneider type lenses (symmetrical/biogon type config) and we can work with distagon type primes, but as yet we have not carried out any precision work with zoom lenses (very hard to achieve).

We are going for side by side arrangement, but for very close inter axial separation we are exploring the use of lateral prisms/mirrors as there is potentially less loss of contrast and the ability to effect greater precision. Our system has to be very precise so that accurate 3d data can be extracted from still or dynamic stereograms. The software we are developing will work with any type of stereo data but if you need to extract 3d geometry from a dynamic scene full 360 degrees you need very precise equipment to make all the 3d pieces fit together into relatively complete models (software can only fix so much). We will (hopefully launch in Q1 2009). The basic rig is much less cumbersome than a traditional mirror based stereo rig and allows exact restitution of the imagery in 6DOF. We have pretty much set out to build a rig in terms of precision and control to enable the best stereograms to be taken. For our 3d/inverse animation system we are still trying find a globally shuttered digital system that is not an industrial CCD camera system (ANY IDEAS?!). The rolling shutter(s) of the RED would make accurate metrology of dynamic 3d scenes very difficult. Never the less we are looking to downgrade the system so that film makers can grab the 3d geometry and texture for simple effects sequences for 3d compositing, rather than a total metrological grade solution.

 

[Benni Diez]

Yeah I think for filmmaking it's not so much about accuracy. It has to be practical and easy to use. And in the end the result has to look convincing.

If you look at some of the stuff they 'dimensionalized', it doesn't look so bad. I've seen a trailer for Nightmare Before Christmas 3D and it was pretty cool. And most of the stuff was crude manual remodelling of the shots. So that's the least Z-depth quality you'll have to deliver and then make it accessible enough for shooters to actually embrace it.

So does your system create 3D geometry for every frame? I imagine some nasty depth wobbling due to noise.

Have you thought about trying something that could be attached to, say, a Red One and deliver data that can later be used in conjunction with the 4k footage?

Because no matter how good cameras you use, it will be hard to get a certain image quality for a nice price.

Also very interesting btw: Nuke has some cool stereo tools. One of them is a plugin that lets you change the actual interocular distance digitally. It uses similar algorithms like those time stretching plugins, so it will depend on the material. But I think one can get away with it in most cases.

I'd really love to see some of your test footage! Or even better, get my hands on it.

 

[Eric Lange]

Yeah I think for filmmaking it's not so much about accuracy. It has to be practical and easy to use. And in the end the result has to look convincing.


So does your system create 3D geometry for every frame? I imagine some nasty depth wobbling due to noise.

Have you thought about trying something that could be attached to, say, a Red One and deliver data that can later be used in conjunction with the 4k footage?

Because no matter how good cameras you use, it will be hard to get a certain image quality for a nice price.

I'd really love to see some of your test footage! Or even better, get my hands on it.

Yes you can extract 3d for every frame. However, we are trying to introduce a workflow where high res medium format cameras are use to capture very high res geometry, at 2 to 1 fps and the lower res high frame rate cameras are used in essence to direct the character rigging, (this is more useful for example for high end gamming content development). I personally would like the system be used in a simple way, just for very accurate 3d recording of as built sets and 3d props etc. to act as a relatively complete 3d record of a film production, as so much gets destroyed the minute shooting is over. That way, the data can be used for spin off media like video games etc, but could also be used in post production for special 3d process shots. The amount of data from four camera pairs for 360 degrees fully extracted into 3d is alarming. We have designed the interface so that stereo sequences can be scrolled through to identify the particular key frames that appear the most useful and then to extract the models from those, rather than extract gigantic amounts of unworkable 3d data. The models created are very clean and compact and because the interface is full stereo VR it has (I believe) one the best manual editing interfaces for dealing with complex 3d data stereo or explicit. [I firmly believe that any automated processes only lives or dies by the quality of their associated manual editing interfaces.] The price point for the software is going to be in the region of $15k, (for a fairly fully featured base package) and the higher end camera positioning platform about the same. [you will be able to mount a RED on it if you want]. However, we are also interested in making inroads into the academic/non profit and cultural domain want to give very good price breaks for the software and also we are designing a downgraded mechanical positioning system that can be used with more simple cameras that will be much more affordable.

The new vapor ware from Jim the DSMC will solve a lot of technical problems and redundancies for us.

Thanks for the input, I’ll definitely get in touch as we are interested in designing the systems around potential users, so it a good time to say what you would like, and what you would like to do and for how much. We also need to find good testers, particularly in Germany as I think there seems to be more of a commercial drive for the adoption of stereo technologies than for example in US right now.


We have paid special attention to clean high quality data and software that is easy to use. The simplicity comes from the best combination of HW and SW design. The software has many features and also include the ability to do parallax resampling to rein in a too larger range of parallax without flattening fine 3d details and texture, (we have patents on some of the basic processes that have been running since 2003).

[My son is giving me dirty looks and wants to play "Heroscape", I'll check in later on, it must be pretty late in Germany]

Cheers,

E.

 

[Benni Diez]

Well personally, as I mentioned in another thread, I'd just love to have a Z-depth channel. Like what the Z-Cam did. But that system seems to be pretty bulky if you want hd quality.
I'm developing a feature film idea that would be very suitable for 3D. But it would also rely on flexibility and portability of the camera.
So maybe a practical side-by-side rig with two Scarlet-like cams with a given distance would be the right thing.
And then being able to extract enough 3D info for compositing and tweaking, and also changing distance and convergence. Add or change depth of field, atmosphere and object distances as I like. Maybe even go as far as changing the structure of objects.
Something that lets me treat the footage like a sculpture rather than an image. And of course makes it a lot easier to combine with CG elements.
I guess reflections and transparency will always be a bitch. So maybe it's better to avoid those things or add them in later.

As for the pricing: I think it would be cool to have something like a plugin for comp programs that gives me a good Z-channel out of stereo footage, with an option to match it with the footage of a third, higher-res camera. Do something like that for After Effects for a fair price and I bet you'll sell them like crazy.
And then the standalone program that goes all the way to creating detailled 3D data and good compatibility with standard 3D software like Max and Maya. Which sounds ok for 15k.

Yeah it's getting late here, I better get some sleep.

 

[Lucas Wilson]

Yeah I think for filmmaking it's not so much about accuracy. It has to be practical and easy to use. And in the end the result has to look convincing.

This is a cool thread...

I was part of a panel of speakers on Monday morning at IBC all talking about the challenges of stereoscopic post-production. Hopefully, I had some bits to add, but the speakers before and after me were awesome. Lots of incredible information about 2D -> 3D, and creation of virtual camera angles in post.

I think the biggest challenge in the industry is going to be education.

For most people, the only difference between "bad" and "good" stereo is a headache, or the feeling that something "just isn't right." Teaching the post industry to quantify what is bad and why is going to be a long process. Because not everything can be fixed in post. A parallel stereo rig shooting something close up will never look right without a *lot* of serious post processing.

Color cubes between an A and B camera require hardware processing to accurately match. Profiling and matching lenses so that each lens's particular set of motion discrepancies are taken into account while setting up the motion rig is hard. From the several stereo rigs I have seen (and I'm sure I haven't seen them all!!!), there are only two companies in the world that have done this correctly - 3ality and Paradise. Eric, I'm sure you'll be the 3rd.

So all that to circle around and say that I think it is absolutely about accuracy... and practicality.

I think U23D was *easily* the best stereoscopic movie ever produced, because the post work was amazing, and the effects were done by VFX houses who know how to do Stereo.

But look at U23D, then look at Hannah. I had a ton of eyestrain during Hannah. So did my daughter... who got a headache halfway through. (it didn't stop her from watching the rest of it, though. )

I am not enough of a stereo expert to be able to look at each one and quantifiably say exactly what was different, and why U2 was seamless for me and Hannah gave me bad eyestrain. How much of it was the rigs, how much was post, and how much was encoding, distribution, and projection? I don't know... but that kind of education is what I think has to happen in the industry before stereo becomes as huge as everyone hopes it will be.

Best,

Lucas
------
VP of Left Eye, EVP of Right Eye.
ASSIMILATE, inc.
LA, CA, USA

 

[Eric Lange]

Well personally, as I mentioned in another thread, I'd just love to have a Z-depth channel.

I guess reflections and transparency will always be a bitch. So maybe it's better to avoid those things or add them in later.


Yeah it's getting late here, I better get some sleep.

Funny you should mention that, if you are out-putting to stereo 3d content rather than monscopic then it won’t be bitch. This is because we have a new type of VR model that uses stereo texturing rather than mono texturing. If you have reflections and transparency then these are stereoscopically “baked” into the model. Our new type of VR model is called a coherently stereo textured model or CSTM. I originally designed this type of model to enable the recording and VR presentation of extremely large and complex surface associated with historic and archeological site as part of PhD. The polygon geometry acts as a 3d substrate for the stereo texture. Due to stereo textural dominance the polygon surfaces are not visible, only the stereo texture. Each vertex acts as what we call a (tri-homologous point) where its apparent position and actual position in 3d space is one and the same. Effectively parallax at the surface of the substrate is reduced to zero leaving a residual parallax surface at the surface of the substrate. This is a very powerful technique if you are using 3d output, more so than 2d. The models only require a fairly sparse number of vertices to effect a very life like representation because of the “perfectly” baked in stereo textures (i.e. you can get away with murder); consequently the rendering is 400X more efficient than conventional methods for effecting the same level of complexity. The rendering of these models also allows a new type of parallax re-sampling once the models are built so that new type of macro and micro parallax can be adjusted real time. If the Video games developers get with stereo then this will be terrific technique for them as it provides extremely realistic models at 1/400th of the compute power.

The use of CSTM’s also permit’s the representation of (you will like this) explosions, miasmas, plasmas, vapors etc. without the need for particle rendering systems (same principal as for reflections, and transparency, fluids water etc.). CSTMs can also be created of completely synthetically rendered/generated CG content, the more complex the better.

{at this point I should be cordially shot for hijacking this thread!}

I’ll personally get in touch and talk some more to find out what you would like to see/have in such a system. We have developed the core capabilities but they can be re-cut or re- jigged into many different types of products so we are always interested in what types of problems people are having with current technologies and methodologies.

 

[Benni Diez]

I absolutely agree. What I meant with accuracy was more the amount and resolution of the depth information that such a post process should provide in order to sell 3D.
It's a bit like cheating with projection mapping to sell depth in a locked off scene.
There's a video on reel-show.tv about Lions 3D. An Imax film that was intended to be 3D, but then shot 2D for butgetary reasons. But then they wanted 3D after all, so they had to painstakingly recreate the depth. The images give you a good idea how detailled the depth-matte needs to be.

There's no doubt that hardware calibration has to be spot on, of course. The synchronizing is where all the perfection is needed. I really had a hard time watching Ghosts Of The Abyss on Imax just because of the mild flickering of the film projection.

As for the rigs: It just depends on what you shoot. For close-up stuff I'd use a mirror rig. But if it's about handheld action they're still too huge and heavy.

 

[cinepost35]

This is a cool thread...

I was part of a panel of speakers on Monday morning at IBC all talking about the challenges of stereoscopic post-production. Hopefully, I had some bits to add, but the speakers before and after me were awesome. Lots of incredible information about 2D -> 3D, and creation of virtual camera angles in post.

I think the biggest challenge in the industry is going to be education.

For most people, the only difference between "bad" and "good" stereo is a headache, or the feeling that something "just isn't right." Teaching the post industry to quantify what is bad and why is going to be a long process. Because not everything can be fixed in post. A parallel stereo rig shooting something close up will never look right without a *lot* of serious post processing.

Color cubes between an A and B camera require hardware processing to accurately match. Profiling and matching lenses so that each lens's particular set of motion discrepancies are taken into account while setting up the motion rig is hard. From the several stereo rigs I have seen (and I'm sure I haven't seen them all!!!), there are only two companies in the world that have done this correctly - 3ality and Paradise. Eric, I'm sure you'll be the 3rd.

So all that to circle around and say that I think it is absolutely about accuracy... and practicality.

I think U23D was *easily* the best stereoscopic movie ever produced, because the post work was amazing, and the effects were done by VFX houses who know how to do Stereo.

But look at U23D, then look at Hannah. I had a ton of eyestrain during Hannah. So did my daughter... who got a headache halfway through. (it didn't stop her from watching the rest of it, though. )

I am not enough of a stereo expert to be able to look at each one and quantifiably say exactly what was different, and why U2 was seamless for me and Hannah gave me bad eyestrain. How much of it was the rigs, how much was post, and how much was encoding, distribution, and projection? I don't know... but that kind of education is what I think has to happen in the industry before stereo becomes as huge as everyone hopes it will be.

Best,

Lucas
------
VP of Left Eye, EVP of Right Eye.
ASSIMILATE, inc.
LA, CA, USA

90% of the cameras used on U2 where the same ones used on Hannah, maybe something different in post process between the two.

 

[Eric Lange]

This is a cool thread...


I think the biggest challenge in the industry is going to be education.

So all that to circle around and say that I think it is absolutely about accuracy... and practicality.

Lucas
------
VP of Left Eye, EVP of Right Eye.
ASSIMILATE, inc.
LA, CA, USA

Luki I really appreciate your post.

We are in essence betting the farm on precision rigs and clean software that is designed with common sense, but as you mention education is extremely important. For our 3d capture techniques we tend to use Biogon (type) lenses so that the radial distortion is kept very low. This means when you play back stereo footage you do not have Y parallax. We further calibrate the radial distortion curves to about 3 micron also for our accurate 3d work. The problem here is that if you use Distagon lenses for a shoot, these will minimally have 2% radial distortion which causes appreciable Y parallax out from the centre of the image. To carry out Bi-cubic re-sampling on a million frames or so just to remove radial distortion is probably something that is not carried out right now. [Y parallax is something that can contribute to stereoscopic fatigue]. At the same time I’m not sure that many cinematographers would be comfortable shooting using special low distortion /biogon/Schneider lenses. I think very large ranges of parallax cause the biggest head aches. We have techniques that allow one to shoot with a relatively wide camera base separation and then to resample these parallaxes into a much more compressed range without compressing the fine 3d detail (our process is unique in this respect). Traditionally you either had to go for safe stereo where it looks flat and devoid of fine 3d rich surface detail, or you had to go for very large ranges of parallax that had you carrying your eyeballs in a glass after the presentation. Our techniques get around this problem by having the capability to independently preserve very fine surface 3d detail. The rigs we have designed also have precision alignment so that the kappa angles of the sensors have rotationally corresponding scan lines, I.e. mechanically produced epi-polar geometry. We are hoping folks will buy rigs from us just to carry out tests to find out what all the permitted variance and tolerance in camera orientation should be (coupled with the appropriate post processes) I.e. use our equipment to determine what the “safety’ standards should be.

I fear though that unless we are able to get through and educate the relevant users, that our technologies may fall through the cracks. My father used to be a production designer on some very big name movies for their time, and I was always appalled at how much beautiful work and craftsmanship was completely destroyed and on the back lot in a matter of days after the shoot. I would love to see our stereoscopic recording and 3d extraction techniques to become a standard film technique and insurance for everything that was physically built on a production, with obvious potential benefits in post, multimedia spin off and so on; but I think also as a permanent record of all the beautiful work that never shows up on film or is really ever seen.

 

[Lucas Wilson]

90% of the cameras used on U2 where the same ones used on Hannah, maybe something different in post process between the two.

I know... most of them were the Pace rigs. Only different was that U2 had a couple of 3ality rigs on it as well.

But the setup and crew were very different, and I know the post processing path was totally different. U2 was done almost exclusively at 3ality, with some final color at Technicolor. Hannah was done at Fotokem.

Lucas

 

[cinepost35]

Well I think 3Ality may have rented also a Hienz rig and I had heard one of the other rigs they had didn't work. I know when you look at the credits it did have some of the same crew on both U2 and Hannah.
I think it must be the post process that was different. 3Ality mostly used Scratch correct? Didn't U2 get printed to 15/70 at Photokem?

 

[Lucas Wilson]

Well I think 3Ality may have rented also a Hienz rig and I had heard one of the other rigs they had didn't work. I know when you look at the credits it did have some of the same crew on both U2 and Hannah.
I think it must be the post process that was different. 3Ality mostly used Scratch correct? Didn't U2 get printed to 15/70 at Photokem?

For U2, 3ality used SCRATCH for conform and dailies. They used their own proprietary in-house tools for stereo correction and finish. Technicolor did the DI and DCP. It was exclusively digital delivery and wasn't printed to anything.

Hannah's conform and DI was on Pablo. DCP was also done at Fotokem.

Lucas