Mary Poppins had more magic than you know
April 7, 2024 9:22 AM Subscribe
The folks at Corridor Crew recently reached back sixty years to recreate a truly wonderous piece of special effects technology that was thought to be long lost.
The video is about the Sodium Vapor Process, if you need some reading first to motivate watching the video. It's an effect sort of like bluescreening, a way to separate the actor from the background. But it's an old fashioned analog technique relying on a special lamp and a prism to split the light.That gives it some nice affordances. (The prism is hard to manufacture, which is why it's not a common effect). The video is great, worth a watch!
posted by Nelson at 9:45 AM on April 7 [9 favorites]
posted by Nelson at 9:45 AM on April 7 [9 favorites]
Gosh wow oh boy oh boy. It's amazing what can be done with off the shelf tech these days.
The original Disney crystals are probably next to the Ark of the Covenant.
posted by Nancy Lebovitz at 9:48 AM on April 7 [2 favorites]
The original Disney crystals are probably next to the Ark of the Covenant.
posted by Nancy Lebovitz at 9:48 AM on April 7 [2 favorites]
I think I only needed to know one thing going in to watch this:
Sodium vapor lights emit an incredibly specific shade of orange, that can be subtracted very easily — whether in astronomy or in special effects — without disrupting orange or other colors.
posted by Callisto Prime at 9:52 AM on April 7 [2 favorites]
Sodium vapor lights emit an incredibly specific shade of orange, that can be subtracted very easily — whether in astronomy or in special effects — without disrupting orange or other colors.
posted by Callisto Prime at 9:52 AM on April 7 [2 favorites]
Mind. Blown.
So cool.
And I immediately start to imagine never having to have suffered through decades of simply awful matting in films and tv, and then I weep.
posted by Thorzdad at 10:08 AM on April 7 [5 favorites]
So cool.
And I immediately start to imagine never having to have suffered through decades of simply awful matting in films and tv, and then I weep.
posted by Thorzdad at 10:08 AM on April 7 [5 favorites]
That was fascinating!
Also, "the original Disney crystals" would be a great username.
posted by Greg_Ace at 11:21 AM on April 7 [3 favorites]
Also, "the original Disney crystals" would be a great username.
posted by Greg_Ace at 11:21 AM on April 7 [3 favorites]
That's really cool! Thanks for sharing the video!
posted by belladonna at 11:24 AM on April 7
posted by belladonna at 11:24 AM on April 7
I totally lack the visual acuity to appreciate the difference, but I LOVE learning about optics - magical stuff. Reminds me of this video on tilt-shift works (and how all the old-timey cameras with the accordion could tilt-shift by default).
posted by McBearclaw at 11:27 AM on April 7
posted by McBearclaw at 11:27 AM on April 7
flabdablet: that was my first thought. Silicon sensors see pretty well into near IR, with pretty good sensitivity up to around 850nm. Removing the hot mirror filter that normally cuts the IR/UV bands is pretty straightforward (at least one company will do it to most models of camera for <$500), and most lenses work well in IR, even if they aren't designed for it.
I think the tricky part would come from the difference in refraction and diffraction between IR and visible (especially IR vs blue): focus would shift, and you would have more diffractive blur in IR. It might be possible to compensate by adjusting the focus and opening the aperture more on the IR camera (assuming DoF doesn't cause problems).
Using a visible narrow band also probably helps humans to set things up, and you could theoretically use narrowband sources for RGB, which would give you very little overlap with the yellow sodium light. Yellow is also nice because it sits near the middle of the large spectral gap between R and G.
posted by reventlov at 11:32 AM on April 7 [1 favorite]
I think the tricky part would come from the difference in refraction and diffraction between IR and visible (especially IR vs blue): focus would shift, and you would have more diffractive blur in IR. It might be possible to compensate by adjusting the focus and opening the aperture more on the IR camera (assuming DoF doesn't cause problems).
Using a visible narrow band also probably helps humans to set things up, and you could theoretically use narrowband sources for RGB, which would give you very little overlap with the yellow sodium light. Yellow is also nice because it sits near the middle of the large spectral gap between R and G.
posted by reventlov at 11:32 AM on April 7 [1 favorite]
Does anybody know why the people in the video keep saying things like "this is the first time in 30 years that anybody is using the sodium vapour process"? Was there some other movie shot about 30 years ago that used the same technique, or are these guys just doing that thing you do when you get old and assume things you saw as a kid must have only happened a few years back while failing to calculate the number of years they've been on the earth into the equation?
posted by sardonyx at 12:28 PM on April 7 [3 favorites]
posted by sardonyx at 12:28 PM on April 7 [3 favorites]
>Was there some other movie shot about 30 years ago that used the same technique...Yes, Dick Tracy, shot in 1989, was the last time a major feature film used the sodium vapor process. So 35 years ago to be precise.
posted by theory at 12:38 PM on April 7 [23 favorites]
Also, "the original Disney crystals" would be a great username.
Weren’t the Original Disney Crystals a plot point in one of the Kingdom Hearts games?
More seriously: cool science, cool application, cool video.
posted by GenjiandProust at 12:39 PM on April 7 [2 favorites]
Weren’t the Original Disney Crystals a plot point in one of the Kingdom Hearts games?
More seriously: cool science, cool application, cool video.
posted by GenjiandProust at 12:39 PM on April 7 [2 favorites]
Yes, IR matting works, works well, and is simple and cheap to do on a large scale. Floodlighting IR on a large rear projection screen makes possible nearly unlimited camera movement and wide field of view. Also, the matte lighting isn’t nearly as fussy to set up as chromakey.
It was done almost twenty years ago on the Las Vegas strip and elsewhere.
posted by lothar at 1:02 PM on April 7 [3 favorites]
It was done almost twenty years ago on the Las Vegas strip and elsewhere.
posted by lothar at 1:02 PM on April 7 [3 favorites]
A little more on the use of the sodium vapor process in Dick Tracy --
Back then blue screens were often the best choice for chroma keying instead of today's more common green screens due to celluloid having peak sensitivity in the blue wavelengths, as well as blue screens having lower luminance than green and thus causing less light spill and color fringing. So, according to cinematographer David Mullen, the sodium vapor process was used in Dick Tracy just for one shot in an opera house that had a blue color scheme (lit by the great Vittorio Storaro).
posted by theory at 1:23 PM on April 7 [7 favorites]
Back then blue screens were often the best choice for chroma keying instead of today's more common green screens due to celluloid having peak sensitivity in the blue wavelengths, as well as blue screens having lower luminance than green and thus causing less light spill and color fringing. So, according to cinematographer David Mullen, the sodium vapor process was used in Dick Tracy just for one shot in an opera house that had a blue color scheme (lit by the great Vittorio Storaro).
posted by theory at 1:23 PM on April 7 [7 favorites]
It's possible that IR wouldn't permit quite as sharp a focus because the wave lengths are longer, but I'm guessing.
posted by Nancy Lebovitz at 1:33 PM on April 7
posted by Nancy Lebovitz at 1:33 PM on April 7
There are a whole number of single light-wave spectrum light sources. I wonder if others could be used to different advantages.
posted by dances_with_sneetches at 2:15 PM on April 7
posted by dances_with_sneetches at 2:15 PM on April 7
As much of a practical effects film buff as I am, I've always wondered how they did these matting effects so well for Mary Poppins, and I just assumed it was manually rotoscoped Disney magic and very labor intensive optical compositing and masking/fritting.
I think I just assumed they did the live action compositing frame-by-frame with a multiplane animation camera instead of using any kind of chromakey and matte printing effect, because I could totally see them doing this with some kind of transparency film print of all the needed frames and just ripping through it with a crew of multiplane animation camera ops.
This was a fun post and I learned something new.
This also makes me wonder if you could do the same thing with modern lighting, but with some kind of structured or tuned LED lights, kind of like the opposite of using the very narrow light spectrum of sodium lights, and instead they used keyed/timed full spectrum lights on multiple channels so that they were indexed to the filming speed so that you "swept" across the RGB spectrum at least once per frame and then generated multiple optical channels to use for matting and compositing to capture full color mattes and keyed frames.
So instead of one static chroma-key color like a green or blue screen or sodium light channel, the background and foreground lights worked together to isolate the full color spectrum of the subject or actors and you could generate multiple matte/fritt shots and automate all that for compositing in post.
Granted it's probably cheaper and easier to just use an immersive live video wall or environment like "The Tank" as used in the Mandalorin.
Making movies is enough of a pain in the ass without having to try to manage and sync 3+1 or more RGB channels just to composite a chromakey shot, heh.
posted by loquacious at 4:01 PM on April 7 [1 favorite]
I think I just assumed they did the live action compositing frame-by-frame with a multiplane animation camera instead of using any kind of chromakey and matte printing effect, because I could totally see them doing this with some kind of transparency film print of all the needed frames and just ripping through it with a crew of multiplane animation camera ops.
This was a fun post and I learned something new.
This also makes me wonder if you could do the same thing with modern lighting, but with some kind of structured or tuned LED lights, kind of like the opposite of using the very narrow light spectrum of sodium lights, and instead they used keyed/timed full spectrum lights on multiple channels so that they were indexed to the filming speed so that you "swept" across the RGB spectrum at least once per frame and then generated multiple optical channels to use for matting and compositing to capture full color mattes and keyed frames.
So instead of one static chroma-key color like a green or blue screen or sodium light channel, the background and foreground lights worked together to isolate the full color spectrum of the subject or actors and you could generate multiple matte/fritt shots and automate all that for compositing in post.
Granted it's probably cheaper and easier to just use an immersive live video wall or environment like "The Tank" as used in the Mandalorin.
Making movies is enough of a pain in the ass without having to try to manage and sync 3+1 or more RGB channels just to composite a chromakey shot, heh.
posted by loquacious at 4:01 PM on April 7 [1 favorite]
I had a vague recollection that, contemporaneous with this Disney technology, the film industry in the USSR was experimenting with Infra-red a.k.a. "hot screen" matting. I confirmed this just now by finding an article in the US SMPTE Journal from September 1963, reporting on one Soviet film stage having an 18 foot by 36 foot "hot screen" lit to a "dull red" color by 9,066 lamps drawing a total of 500 Kilowatts ( ! ) The camera was fitted with a beam splitting prism to direct infrared light to one film, and the normal color image to another. I imagine that not much sophisticated filtering was needed in that prism, given the relative insensitivity of the color film to infrared, and the incredible infrared brightness of the backdrop compared to the normally-lit foreground.
There was supposedly also an experimental variation that tried to eliminate the need for a beam splitting prism entirely, by coating the back of the color film with an infra-red sensitive emulsion. This was developed to a black and white image, which was either physically transferred to another carrier strip of film after development, or scanned with an image scanner and then allowed to bleach out along with the silver image of the front-side color image, as a normal part of the post-color-development bleaching process of color positive development.
posted by ReferenceDesk at 4:47 PM on April 7 [7 favorites]
There was supposedly also an experimental variation that tried to eliminate the need for a beam splitting prism entirely, by coating the back of the color film with an infra-red sensitive emulsion. This was developed to a black and white image, which was either physically transferred to another carrier strip of film after development, or scanned with an image scanner and then allowed to bleach out along with the silver image of the front-side color image, as a normal part of the post-color-development bleaching process of color positive development.
posted by ReferenceDesk at 4:47 PM on April 7 [7 favorites]
some kind of structured or tuned LED lights, kind of like the opposite of using the very narrow light spectrum of sodium lights, and instead they used keyed/timed full spectrum lights on multiple channels so that they were indexed to the filming speed so that you "swept" across the RGB spectrum at least once per frame and then generated multiple optical channels to use for matting and compositing to capture full color mattes and keyed frames
Super color fragile, is trick LEDs a better process?
posted by flabdablet at 9:29 PM on April 7 [20 favorites]
Super color fragile, is trick LEDs a better process?
posted by flabdablet at 9:29 PM on April 7 [20 favorites]
Was recently rewatching Lost, and there were some chromakey scenes that looked truly awful, mostly in the first or second season (2004-2006) and usually faces against sunsets or other dramatic backgrounds. They tried to cover it with heavy feathering but it still looks like poo. And that was a big budget smash hit production. Even in 2024 I thought this would be a solved problem but every now and then I catch a modern TV show or whatever leaning on that feathering to hide crimes.
posted by Rhomboid at 11:24 PM on April 7
posted by Rhomboid at 11:24 PM on April 7
This is amazing! I had wondered why old school compositing sometimes looks so great. I wondered if they used this for the overgrown Washington DC at the end of Logan's Run, but apparently that was matte painting.
posted by TheophileEscargot at 11:54 PM on April 7 [1 favorite]
posted by TheophileEscargot at 11:54 PM on April 7 [1 favorite]
I just want to say that the green and blue fill/back lights on the initial interview segments were absolutely perfect environmental storytelling. No notes.
posted by rum-soaked space hobo at 8:31 AM on April 8
posted by rum-soaked space hobo at 8:31 AM on April 8
Was recently rewatching Lost, and there were some chromakey scenes that looked truly awful, mostly in the first or second season (2004-2006) and usually faces against sunsets or other dramatic backgrounds.
There's a lot of late '90s / early '00s effects that just don't hold up on modern TVs. CRTs and 720p LCD/plasma displays hid a lot of sins.
posted by nathan_teske at 11:15 AM on April 8
There's a lot of late '90s / early '00s effects that just don't hold up on modern TVs. CRTs and 720p LCD/plasma displays hid a lot of sins.
posted by nathan_teske at 11:15 AM on April 8
IR would be tricky because there are so random sources of heat on a movie set, FWIW.
posted by Pitachu at 3:24 PM on April 8
posted by Pitachu at 3:24 PM on April 8
On further reflection,
keyed/timed full spectrum lights on multiple channels ... "swept" across the RGB spectrum at least once per frame
is an interesting idea. Given that LEDs can be switched between full and zero brightness way faster than commonly used movie camera frame rates, it should now be feasible to use time-division rather than frequency-division multiplexing for lighting the backing screen vs the foreground action.
If a translucent screen's backlight and all the foreground lighting were alternated at say 240Hz, and the camera shutter synchronized to that so that every other frame captured only the backlit silhouettes, then it seems to me that the image processing required to recover very cleanly separated 24Hz matte and foreground frame streams from the resulting pair of interleaved 120Hz streams would not need to involve spectrum sweeping or indeed any kind of color-sensitive processing, just downconversion to 24Hz.
I'd expect the motion blur inherent in generating each 24Hz frame from five 120Hz frames to be more than enough to bury any residual motion fringing artifacts arising from the inherent 4ms offset between matte and foreground streams.
posted by flabdablet at 9:05 PM on April 9 [2 favorites]
keyed/timed full spectrum lights on multiple channels ... "swept" across the RGB spectrum at least once per frame
is an interesting idea. Given that LEDs can be switched between full and zero brightness way faster than commonly used movie camera frame rates, it should now be feasible to use time-division rather than frequency-division multiplexing for lighting the backing screen vs the foreground action.
If a translucent screen's backlight and all the foreground lighting were alternated at say 240Hz, and the camera shutter synchronized to that so that every other frame captured only the backlit silhouettes, then it seems to me that the image processing required to recover very cleanly separated 24Hz matte and foreground frame streams from the resulting pair of interleaved 120Hz streams would not need to involve spectrum sweeping or indeed any kind of color-sensitive processing, just downconversion to 24Hz.
I'd expect the motion blur inherent in generating each 24Hz frame from five 120Hz frames to be more than enough to bury any residual motion fringing artifacts arising from the inherent 4ms offset between matte and foreground streams.
posted by flabdablet at 9:05 PM on April 9 [2 favorites]
Mod note: This Supercalifragilisticexpialidocious post has been added to the sidebar and Best Of blog!
posted by Brandon Blatcher (staff) at 6:57 AM on April 10 [1 favorite]
posted by Brandon Blatcher (staff) at 6:57 AM on April 10 [1 favorite]
Their self-described janky setup was really trivial to put together. The throughput would be awful compared to the original filter embedded within the beam splitter, but that can't be the only reason that nobody used the sodium technique. What is the pain point? Is setting up the lighting too expensive?
posted by Prof. Danger at 9:19 PM on April 10
posted by Prof. Danger at 9:19 PM on April 10
« Older The comics legend lurking in a Sunderland basement | When things break Newer »
I wonder whether something similar could be done using infrared or ultraviolet instead of monochromatic sodium vapour so that the mask lighting entirely avoids the visible colour spectrum. IR filters are cheap and readily available.
posted by flabdablet at 9:44 AM on April 7 [3 favorites]