|
The Magic of Mary Poppins Rediscovered
Fascinating and fun video on rediscovering the technique Disney invented to produce impeccable matting in Mary Poppins.
|
Wow. Awesome.
|
I’m an unashamed special effects geek and science nerd. I knew about the sodium vapor technique, and how effective it was in Mary Poppins. But…
I had no idea that they’d lost the tech to do it. I’m really stoked that someone managed to realize the technique with new tech - and off-the-shelf tech at that! Thanks!!! |
Apparently, the sodium-vapor technique was last used in Dick Tracy.
|
very cool.
|
This is very interesting, but my inner-asshole is compelled to rain on the parade a bit.
The technique wasn’t lost, it died |
Well, if disney was the only one doing it, did they lose the special prism in storage or under a desk somewhere, or did they decide greenscreen was easier? From these guys (time spent to feather, trim, masking) it doesn't sound like it.
Did disney own copyright so no one else could, or was this some kind of top secret thing? you are our film expert so opine away. |
Can't be copyright - it wouldn't have expired yet. Maybe a patent on the prism cubes that were too hard to make.
Corporate copyrights last forever (95 years) to maximize shareholder value. |
I believe they actually lost the custom prisms.
|
|
To me, calling a technology “lost” strongly implies the process involves a mystery greater than a missing part we can create a facsimile of.
|
|
The reason they were having so much difficulty with the greenscreen in the video was because they contrived a job filled with all the things greenscreen handles poorly. No one actually does this, so in practice it’s never as difficult as it was in the video. |
|
subego, I'd expected you to be more excited by this. Yes, you can manually avoid the greenscreen limitations. Or switch to bluescreen for a scene with green present. But that's accepting a handicap. The narrower the light band used for matting, the fewer conflicts you run into. And you don't have to spend time & planning avoiding all your handicaps. Reading your replies, it sounds kind of like "but the wheelchair is comfortable". Perhaps your cats will eventually notice the portal to Mars in the back of the fridge. So that's why the adults let so much old food turn gross - to keep a portal hidden. |
To shoot a sodium vapor shot requires:
1) A duplicate camera. 2) A beamsplitter. If the camera lenses are expensive, so is this. 3) Specialized lights which I can’t use for anything else. Also, the beamsplitter cuts the light each camera receives in half. If I want to compensate for that I need twice the output from all my lights. To use an extreme example, I’ve done bluescreen with a roll of colored paper lit by the sun. Being confined to the wheelchair has its advantages. :) The scientist in the video has a grounded perspective. At the end he’s asked what we get out of resurrecting the process. He doesn’t answer it will become a resurgent technique, but rather it will get used as food for AI greenscreen software. |
I've done manual clipping paths in photoshop. Even the new photoshop methods are not the magic the ads show (fine hair, transparent fabrics, water glasses, reflections...). If it turned out that ye ancient photoshop 3 had a way to do smooth silhouettes without fuss I'd be on board. Even if it involved messing with *channels* ugh.
|
I don't necessarily disagree with the conclusion - if you've done 100 video shoots, that's one hundred more than I have. But regarding the details ...
Why does a custom tool need to be used for something else to be a good value? I don't expect my coax stripper to be useful in the garden. A pipe wrench is mostly useful for pipes. Use on burglars is not recommended - they may sue, and end up owning the pipe wrench. If I buy a kitchen sink to wash dishes -- must it do double duty for something else, or it's a bad deal? ps - what else does one do with a kitchen sink? Wall art? Meth cooking? |
|
I consider the sodium vapor process “lost” in the way the tech to “manufacture a Saturn V” is “lost”. The jigs, production lines, and most critically the personal expertise needed to build those behemoths no longer exists. They were literally hand-crafted, particularly the main engines for the first stage, the Rocketdyne F-1. No two of those engines were identical; the fuel/oxidizer injection plates were HAND DRILLED, each and every one.
Not being able to just “turn on” a production line for the Saturn V is not a bad thing. The art and science of boosters grew massively after the end of the Apollo Program. SpaceX’s Raptor engine is so much more efficient and powerful per pound than the F-1 that recreating a flyable F-1 engine wouldn’t be worth trying. Similarly, discovering that one doesn’t really “need” that custom and expensive prism to do the technique also demonstrates that more modern tech can provide more efficient, more affordable, and more flexible ways to use sodium vapor direct mattes. Sometimes losing a specific tool simply makes you create a better tool. Like in this case. |
However, I subscribe to the philosophy a tool’s defining trait is the frequency it does a good job solving problems. The higher the frequency the higher the value of the tool. Tools that successfully merge duties will have higher frequencies, so my outlook favors them. Of course, frequency isn’t the only thing that gives a tool value, but a low frequency tool has to bring more to the table than a high frequency one. |
Along those lines, and the same lines Laminar mentioned, every time the focal length needs to change, the system will need realignment. I presume this is one of the reasons the Mouse went through the trouble of making the custom beamsplitter: so it could go behind the lens and not need any fussing after a lens swap. Further, with the Mouse version it’s possible to use a zoom, which means focal length can be adjusted with a simple twist of the barrel. |
If it were me, I’d be tripping over myself to downplay it. |
well you could still bill it as a long complicated process requiring lots of filters and outlining, right? right?
|
Which process?
|
Oh! For the sodium vapor.
I’d bill essentially nothing for the actual post-processing, but I’d legit charge an arm and a leg to get the shot. Similarly, I question whether doing the greenscreen clown bride was appropriately budgeted. How many billable hours did they put into the post-processing? It needs more than a hundred. They acted like if someone gave them the clown bride greenscreen shot they would say it’s impossible. I have this sneaking suspicion the real answer would be “pay me enough and we can make it work”. As an aside, I’d pitch we lose the real veil and make a CG one… unless that’s not realistic enough for a clown bride on Mars. ;) |
heck, get AI to do everyth---wait our jobs! come back!
(I used regenerative fill today to extend a photo to better fit. I feel dirty.) |
They still need humans to count fingers.
|
So you get simple zoom, quick-mount of the cams, and a 2-cam remote for shutter and pic adjustments. The combiner body won't be cheap, but probably way cheaper than Disney Cubes. The cost of a cube should include all the failed attempts before they got another working one. Anyone know how many attempts Disney made, to end up with 3 acceptable Cubes? |
We can throw the exact same scenario at both processes as an experiment and a demonstration, but if we actually wanted to make the greenscreen version there would be alterations made to accommodate the medium. Take the dress with the green dots for example. Probably better to have those be another color and have a separate effect to change them green rather than pile it in with all the other greenscreen problems. |
No matter what we do, this is going to degrade the image. In other words, for all of its faults, because there’s no intermediate optics stage involved the greenscreen shot will always be cleaner. The less the optics stage degrades the image, the more it will cost to develop and manufacture. That’s what likely kills it. We need to get Zeiss on the blower, and they’re not going to come cheap. |
The ideal solution would be a custom CCD camera sensor. Besides the usual RGB subpixels, it would have a 4th subpixel, only sensitive to 589 nm. So you can yellow-screen with a single cam, with the extra subpixel being recorded in the alpha channel. No extra lens elements or beamsplitter, and you have the masking data available for every pixel.
This would be more precise than aligning two cameras after the fact, and all the other negatives go away too. No loss of brightness, no extra optical elements, use of existing lenses. Use of existing camera bodies too - a manufacturer would just install the special chip. Minimal firmware changes, just make sure the extra data is recorded to the alpha channel. The only remaining negative is the cost of course. Custom silicon can't be cheap, unless it catches on big for production cameras. |
The other issue is volatility of sensor tech. Top cameras have 12K pixels horizontal now.
|
|
The first shot of a movie I watched tonight was the protagonist watching a storm roll in. The second shot was behind the protagonist, with the storm composited in. It starts raining, and the shot-reverse shot happens a few more times.
In the first shot, they’ve got a fan blowing around little wisps of the protagonist’s hair. For the second, they didn’t even try. The rain hadn’t started yet but they used a shot where his hair was already wet enough it was anchored in place. That said, I wouldn’t say it was really noticeable per se. For various reasons, this thread among them, the moment I realized it was a composite I started analyzing the hair. That’s not how most people watch movies. I rely on that to sone extent. If I’m confident the audience won’t be looking at where the problem is, sometimes I’ll just let lit slide. In this particular case the storm is what’s compelling about the shot. All the composite needs to do is not draw attention to itself. It didn’t. |
Good point. But sometimes there's things like Henry Cavill's mustache which mean that someone dropped the ball on estimating what the audience will notice.
A project at work is considering a brand update. New fonts were proposed. The one the team is leaning towards would be barely noticeable as different to the non-design layperson. My thought is why bother changing (all the files, all the art, reprinting materials) for such a small change, for changes sake only. |
|
It is one of the top places to look at Henry Cavill, true.
|
In that case though, I have to admit the movie did such a good job of being flaming garbage it distracted me from the lip.
I’ve missed a lot of his stuff. I think my faves are the Enola Holmes movies. I’m excited about his 40K project. |
All times are GMT -4. The time now is 11:05 PM. |
|
Copyright © 2005-2007 MacNN. All rights reserved.
Powered by vBulletin® Version 3.8.8
Copyright ©2000 - 2024, vBulletin Solutions, Inc.