The Hobbit at 48fps...

[Zeb B]

If Space Odyssey 2001 was shot today, and considering its long stable shots, I think it can greatly benefit from 48fps 3D.

you're trolling . . . right?

 

[Ian McKenzie]

This thread has thrown up a very interesting paradox.
On the one hand we have all the praise for what Jim Jannard and Red have done with their 4k cameras that is, justifiably imo, evident thruout these forums.
On the other hand we have the dislike and criticism of what happens when someone (Peter Jackson in this case) uses those cameras in a manner for which they were specifically designed and intended to be used, ie 4k and higher frame rates than 24fps.
Imagine the complaints if The Hobbit was not just shot at 4k / 48fps but shot and displayed at 4k / 48 fps or even 60fps, which the Dragon sensor can accomplish.
How much more fake and "video" looking would the sets, make-up, motion etc look then?
There is a very obvious and interesting ( to me at least ) contradiction here.
Do those who are unhappy with the way Peter Jackson has used his Red cameras intend to only shoot at 1080P / 24 fps if and when they purchase or rent Red equipment?
Is 1080P enough after all, despite what Jim Jannard claims?
For some people, praising the camera and damning the intended (intended by the filmaker and the manufacturer) results from those cameras is obviously fine !
Interesting.

 

[António Fagundes]

This thread has thrown up a very interesting paradox.
On the one hand we have all the praise for what Jim Jannard and Red have done with their 4k cameras that is, justifiably imo, evident thruout these forums.
On the other hand we have the dislike and criticism of what happens when someone (Peter Jackson in this case) uses those cameras in a manner for which they were specifically designed and intended to be used, ie 4k and higher frame rates than 24fps.
Imagine the complaints if The Hobbit was not just shot at 4k / 48fps but shot and displayed at 4k / 48 fps or even 60fps, which the Dragon sensor can accomplish.
How much more fake and "video" looking would the sets, make-up, motion etc look then?
There is a very obvious and interesting ( to me at least ) contradiction here.
Do those who are unhappy with the way Peter Jackson has used his Red cameras intend to only shoot at 1080P / 24 fps if and when they purchase or rent Red equipment?
Is 1080P enough after all, despite what Jim Jannard claims?
For some people, praising the camera and damning the intended (intended by the filmaker and the manufacturer) results from those cameras is obviously fine !
Interesting.

The problem with this is that its a one size fits all kind of game. And film making is anything but that. If everything now be shot at 5K with 21 stops of DR and at 48/60 fps. I say, welcome to the flat factory. Every film would look/feel the same. Flat. Not flat in terms of shadow/highlight, but flat in terms of dramatic quality in the image.

Lighting is not about how much you can see in dynamic range or in detail, movies are not editorials, sure if that is an option is great to have it, but to assume everything should be made that way is very narrow minded and is clearly drinking a very dull kool aid.

 

[Ian McKenzie]

The points you make are good, valid ones Antonio and yet they don't make any difference to the essential point of the paradox. No matter what style of filmaking a director or DP chooses to use with their Red cameras they are going to be faced with the issues mentioned above. One reason I made the post was because of the Dragon thread on this forum, with all its good news about increased DR etc, which will show even more detail, in both highlights and dark scenes.
Yet some do not like seeing that extra detail etc, to them its is not "filmic".
In other words, Dragon, and Red in general, make things worse, not better !
Personally, I think they are wrong, but they are certainly as entitled to their opinion as I am to mine.
This thread is where we are hearing for the first time (on these forums) some criticisms from Red users and fans who don't like the results of using Red cameras and their ever better sensors etc.
And that I find is interesting. As will be the different ways in which filmakers address such concerns, as you point to in your post. One of the many benefits of Red and the raw workflow is that they do offer filmakers many more artistic options than exist using conventional film.
And that will lead to much debate and discussion, and many differences of opinion.
Interesting times indeed.

 

[kenneth gruyaert]

There will be no need for RED if 24fps remains the standard, it does not take high end technology to record 8k at 24fps and i am sure that very affordable camera's will be capable of this very soon. What exactly will be the selling point for RED camera's?
Personally i think 24fps completely destroys the image quality and i cannot wait for it to go away, even 60fps is not anywhere near good enough so there is plenty of reason for RED to exist in the foreseeable future. I am just curious what reason all the high fps haters see.

 

[Gavin Greenwalt]

There will be no need for RED if 24fps remains the standard, it does not take high end technology to record 8k at 24fps and i am sure that very affordable camera's will be capable of this very soon. What exactly will be the selling point for RED camera's?

Lightfield, HDR and chip size.

Dragon looks plenty sensitive per mm of surface area. Dragon has plenty of resolution to the point that I'm seeing 2px worth of chromatic aberration at least in those shots. Dragon is starting to just capture defective lenses really really well.

Next move will inevitably be into how do you use that resolution for new workflows. With 20 stops of dynamic range you're nearing the point where you almost can't screw up the exposure too badly as long as you shoot F/4ish. I imagine the next chip will have 22+ stops of dynamic range easily and 4000 ISO. That's certainly "good enough" for just about any situation short of a welding torch with the slightest of auto-exposure adjustment.

Add lightfield and you can choose your focus and depth of field in post. Let Autoiris handle exposure to keep the image within the goalposts and then process the lightfield so that you get a constant depth of field.

Add 120fps to the mix and you can shoot a 360* shutter and you could extract a 180* shutter (or less) at 24,30,48 or 60 fps depending on your deliverable from one clip.

And with the lightfield data + motion blur free frames you're able to extract a point cloud for 3D meshing.

Add in a 617 sized sensor and the lightfield and you can get all of the "normal" interocular distances from one frame for stereo.

 

[kenneth gruyaert]

Lightfield, HDR and chip size.

Chip size does not seem like an argument since there are very cheap aps-c camera's and full frame camera's like the nikon D800 are reasonably affordable. Maybe you meant RED will focus on medium format only in the future?

HDR will be easy enough to implement in the low cost camera's too and some time in the future better sensors will support it natively as you point out.

Add lightfield and you can choose your focus and depth of field in post. Let Autoiris handle exposure to keep the image within the goalposts and then process the lightfield so that you get a constant depth of field.

Add 120fps to the mix and you can shoot a 360* shutter and you could extract a 180* shutter (or less) at 24,30,48 or 60 fps depending on your deliverable from one clip.

And with the lightfield data + motion blur free frames you're able to extract a point cloud for 3D meshing.

Add in a 617 sized sensor and the lightfield and you can get all of the "normal" interocular distances from one frame for stereo.

I am not sure what you mean by lightfield? The technique used by lytro is a dead end and i would not expect any real light field sensor anytime soon. Limited depth of field is a problem that is here to stay i am afraid, for me it is the biggest obstacle to really high quality capture and i don't see a solution for it in the next decade at least.

To eliminate motion blur while capturing all available light we will need much more than 120fps. I think eventually at least 4000fps will be used. For a panning speed of one frame width per two seconds this gives 1 pixel width worth of blurring with 360deg capture so it should be near perfect while at faster panning speeds i suspect our eyes probably would not be able to get sharp enough focus on the moving objects to benefit from higher framerates.

 

[Gavin Greenwalt]

The technique used by lytro is a dead end and i would not expect any real light field sensor anytime soon.

You might as well say "I would not expect any usable digital sensor anytime soon." Lightfield capture is today where digital cinema was maybe 10 years ago. And yes I mean RED will focus probably on Medium format (or larger like 617) in the future.

If RED wants to stay at the cutting edge they're going to have to move into computation photography more and more. Part of that will probably involve large sensors to deliver single lens/chip Stereo.

 

[kenneth gruyaert]

Diffraction prevents lytro's technique from ever producing high quality images. Effectively the way lytro works means that the images are computed from a bunch of images taken with tiny aperture camera's which are usually all focused on the wrong depth. No matter how many images you combine taken with the aperture set at F/64 you will still have the same diffraction limited resolution, if you use a much larger aperture lens you might have F/16 equivalent aperture sections but those images will be badly out of focus for most scenes.

Also you mention that it would be possible to extract 3D from a single lens on a large format sensor but it is the aperture of the lens that determines the range of positions that the scene is viewed from and large format lenses do not have larger apertures than full frame lenses. To have real usable results from lytro's technique would require something like a full frame 35mm F/0.5 equivalent lens giving an aperture of 70mm thus covering the interocular distance of most humans, this would likely be at least a 140mm F/2 lens on a 144x96mm sensor to allow the optical quality to be reasonable with the aperture wide open so i don't see this being a realistic solution in the future both because of the size and cost.

Instead of this I think arrays of large numbers of small camera's is a much more realistic possibility for computational photography. This would give a very large virtual aperture to allow for a large amount of motion parallax. The light from the multiple camera's could be accumulated in processing just like is done in lytro's technique, so for example 64 tiny F11 equivalent lenses can produce the low light quality of a single F1.4 lens with the depth of field of F11, it will also be possible to let all the lenses cycle through the depth range to capture sharp focus on the whole scene. I am very much looking foward to computational photography just like you are but just not using lytro's technique.

 

[Nikhil Kamkolkar]

The amazing thing to note in all this is that there are asks for RED to deliver an easier post workflow (which is a very practical sort of a request, and one that's a personal interest of mine), right alongside hopes expressed for RED to deliver the cutting-edge acquisition technology of the future.

Which other company is receiving requests for incremental improvements to its existing, shipping products, even as it is considered the leader for delivering the acquisition technologies of the future?

Bravo, RED!

 

[BRANDON JAMESON]

Diffraction prevents lytro's technique from ever producing high quality images. Effectively the way lytro works means that the images are computed from a bunch of images taken with tiny aperture camera's which are usually all focused on the wrong depth. No matter how many images you combine taken with the aperture set at F/64 you will still have the same diffraction limited resolution, if you use a much larger aperture lens you might have F/16 equivalent aperture sections but those images will be badly out of focus for most scenes.

Also you mention that it would be possible to extract 3D from a single lens on a large format sensor but it is the aperture of the lens that determines the range of positions that the scene is viewed from and large format lenses do not have larger apertures than full frame lenses. To have real usable results from lytro's technique would require something like a full frame 35mm F/0.5 equivalent lens giving an aperture of 70mm thus covering the interocular distance of most humans, this would likely be at least a 140mm F/2 lens on a 144x96mm sensor to allow the optical quality to be reasonable with the aperture wide open so i don't see this being a realistic solution in the future both because of the size and cost.

Instead of this I think arrays of large numbers of small camera's is a much more realistic possibility for computational photography. This would give a very large virtual aperture to allow for a large amount of motion parallax. The light from the multiple camera's could be accumulated in processing just like is done in lytro's technique, so for example 64 tiny F11 equivalent lenses can produce the low light quality of a single F1.4 lens with the depth of field of F11, it will also be possible to let all the lenses cycle through the depth range to capture sharp focus on the whole scene. I am very much looking foward to computational photography just like you are but just not using lytro's technique.

We bought a $ 500. LYTRO ‑ 11‑megaray Light Field Camera, 16gb , based on its manufacturer's unique promises and abilities .....and it is almost completely useless for ANYTHING!.
The images it produces are smaller than a postage stamp, they can only be viewed in a Lytro web-based environment and the rez must be less than 25dpi.

So, with those specs, who cares or can even TELL what amount of post focus shifting is possible. #expensivenonsense

If anyone wants to buy it for 400, youre welcome to it!!!

 

[Marc Wielage]

Getting back to the subject of The Hobbit, there's a terrific article about the workflow and editing of the movie on the Definition Magazine website:

http://www.definitionmagazine.com/journal/2013/2/19/winning-the-frame-game.html

Noted editor Jacob Olssen has some interesting things to say, particularly given that they not only didn't initially edit the movie at 48fps, they also didn't cut it in 3D or 4K:

"For the offline edit we worked in 2D at 24fps with Avid DNxHD36 media. The Avid media is stored on an Avid ISIS 7000 with 116TB of mirrored storage giving 58TB of active space, this contains all the footage so far captured for the three Movies."

"We did not do the DI at 4K at all. This was primarily due to cost and time for the VFX shots if they had been done at 4K (it is four times the data of 2K). The DI was conformed at Park Road Post using an array of SGO Mistikas, it was all 2K and was primarily done at 48fps 3D with the other versions, 24fps 3D and 24fps 2D derived from that. "

The amount of footage they've shot for the three movies is staggering:

"48fps was shot on set, for each eye as it was a 3D shoot. So we were effectively generating 96fps of data. After all the shooting days we had over 50,000 clips, 25,000 per eye. (But each clip might contain several ‘normal’ takes, as often the cameras were not cut between takes. 2,200 hours of footage shot, 1,100 x 2 eyes (24 million feet of film equivalent)! The largest shoot day was 11 hours of footage per eye. These figures are for all three movies combined."

Note that they used 13 Avid Media Composers (12 on Windows, 1 on Mac) to actually cut the movie. Quite an incredibly-complicated production... and it's still going.

 

[Zhang Ming]

48FPS is not available in China...But I can shot some 48FPS 5K S3D clips at home with my Epics and a full set of P+S 3D Rig and optimo3D Lenses to gether with CMotions now, hahaha. I will make my own indie movie in 3d and 4K and 48FPS soon. Hoping I can get the power of dragon then. I must say, the 48FPS S3D 4K edit thing isn't that horrible once you found the trick, but you do have to use multi-software and huge fast RAID to do that, I spend two weeks and tried different softwares and finally get done...some timecode management work also needed.

 

[paulherrin]

zhang, i'm curious what you found to best do the trick