Slashdot Mirror
Delivering 8K VFX Shots For the Dark Knight
agent4256 writes "Barbara Robertson over at Studio Daily put forth this article featuring the technical background for the production of The Dark Knight. With most of the film shot with IMAX cameras (producing a theoretical resolution of 18k), the studios could not handle the size. Instead, they cut the resolution by more than half, down to 8K, the maximum resolution for scanned film. 'A single 8K frame requires 200 MB of data,' Franklin says. 'So we had to upgrade our whole infrastructure. We needed faster network speeds to move data around, massively beefed up servers, and — the most important thing — a new compositing solution.' To give you an idea of how far technology has taken us: 'In 1999, when we worked on Pitch Black [released in 2000], we needed to access 2 TB of data,' Franklin says. 'This show used over 100 TB of data.'"
Except Batman doesn't inexplicably throw thousands of Robins at Catwoman after the Joker does something bad, while he sits back doing nothing
Yeah, no, just no. That's idiotic and is looking for deeper meaning then the meaning that is there. Have you seen the movie?
*****SPOILER ALERT***********
1. Harvey Dent attempts to torture a captured underling to get information out of him, Batman stops this, pointing out he's not going to get anything useful out of him. It was russian roulette torture, not waterboarding, but the connections should be obvious
2. Some city-wide cell-phone based surveillance system is set up by batman, and while it does work the movie makes the point that batman can't be trusted with it, he gives it to the CEO of Wayne enterprises and it gets destroyed right after the joker is caught. Again, they don't actually call it the patriot act, but the parallels are not easy to miss. Bush isn't giving the patriot act to France with the string that they destroy it once osama is caught.
3. While Batman does operate outside the law to get things done, he doesn't make that excuse to duck punishment. At the end, he actually takes on blame that shouldn't be his.
4. Batman uses his own money to fund his fight against the joker, wheras Bush spends my tax money and gives his friends tax breaks.
5. Batman refuses to kill villians and instead turns them over to the justice system. Bush attempts to kill terrorist sympathizers, and refuses to give terror suspects due process.
The reference you quote does make it clear, but you've drawn the wrong conclusion:
> 5.6K: 5616x4096; A full 5.6K was actually...
> 8K: 8192x6144; approximately ....
Thus 8K is 8192 pixels wide (not lines per frame) and 6144 pixels high. We commonly also use 2K's (2048 x 1501), 4K's (4096 x 3002), etc.
Also note that the digital professional cinema (not HDTV) industry (the world of DCI) also always uses image width rather than height to define resolutions (2K = 2048 x 1080, 4K = 4096 x 2160).
[/me = Technical Director on several digital 3D Imax films back through the late '90's -- these Hollywood guys are just now discovering stuff the rest of us have known for ages]
No. Well...maybe. Actually, yes. It really just depends.
-A single 8K frame requires 200 MB of data.
-The Dark Knight is officially listed at 2hrs 30 minutes (150 minutes= 9000seconds)
-Total usage 100 TB (5 frames a Gig, 5120 per T, 512,000+ frames)
Minimal frame rate is ~24/s.
200 MB/frame x 9,000 sec/movie x 24 frames/second = 43200000 MB=42187.5 GB = 41.2 TB.
If the frame rate was 60 frames/second then that would be the whole film (no retakes, extras, bloopers etc).
I never realized the sheer amount of compression that is going on between the raw footage and getting it into a DVD.
2k, 4k and 8k, when referred to film, are the horizontal resolution.
720p and 1080p when referred to TV sizes, refer to the vertical resolution.
Look at the image in the middle with the coloured blocks
http://en.wikipedia.org/wiki/Digital_cinematography [wikipedia.org]
It shows 2K being just slightly larger than 1080p. If 2k referred to the vertical size then it would cover 4 times the area
Back in the mists of time, I wrote the database for the content management system that Lucas used on Star Wars I (the Phantom Menace). For reasons I won't go into, it was called 'Cakes', but ILM rebranded it internally as Media-DB.
At the peak of filming, it was coping with 40 DTF tapes/day being ingested. A DTF held 120GB back then (I think), and they were filming for ~3 months. At the same time as ingesting, it had to stream low-res proxies of all the footage to multiple destinations (some local, some not), and deliver high-res frames across the internal network to the animators etc.
Now, I doubt it was doing 40 tapes/day solidly - it'd depend on filming, but even taking 20 tapes/day, over 3 months that comes to ~160TB (assuming a 22-working-day month).
I do have fond memories of doing the James Bond intro-sequence (The world is not enough) with Smoke & Mirrors in London. When there were thousands of frames of nearly-naked highly-attractive women having oil poured all over their bodies, the visualisation tools became... significantly more advanced at a rapid rate :-)
Simon.
Physicists get Hadrons!
We at the StereoLab in the National Film Board of Canada have an infrastructure set up specifically to manage a number of simultaneous 3D productions, several in "Large Format" (i.e. Imax) resolutions and the rest in various HD and 35mm formats. It's been to make over a dozen 3D digital films in the last few years or so.
In practice we use about an equal mix of internal data server drives, SAN, NAS, and a pool of bare drives with a stack of empty shells. Often people drop a drive in a shell and attach it (via eSATA, FW800 or USB in that order of preference) to whatever machine they need it on, because it reduces network load. This technique works especially well for intermediate data that is output, reinput, and then discarded.
No. Well...maybe. Actually, yes. It really just depends.
Film has a resolution, even though it isn't in the form of nice sharp-edged pixels. It's a question of how close together two objects can be and still be distinguished -- the distance is called the circle of confusion, within which the two objects are not fully distinct. Lenses, film, and printing process all play a role in the resolution of the final product. For test work, one usually uses a printed image with a very fine array of slowly converging lines, and you look for how close together the lines can get before they become indistinct. As a result, the number of (distinguishable) lines you can fit on the film is the natural way to measure its resolution. So film really does have "lines" and though they're not quite the same as in a digital system, they're remarkably close.
(Be aware there's a factor of two in there for Nyquist; a 1000 pixel wide display can only show 500 lines, obviously, and the same effect applies to analog systems.)
Of course, with better digital sensors (ie lots of megapixels), the lens quality becomes the limiting factor, and it would again make sense to speak of the imaging system in terms of lines of resolution rather than megapixels. There's a reason cheap cell phone cameras don't produce as sharp an image as a real camera with a good lens; if you want to measure the quality of the entire imaging system, you end up back with old-fashioned analog lines of resolution as one of the fundamental metrics. (Of course, there are plenty of other attributes, like various forms of noise and distortion.) If you read a good review of a digital camera, they'll point it at a test piece and measure available lines of resolution, just as they would for film.