Note: This article is the "Director's Uncut Version" of an article in the Creative Cow Magazine's Film Values Issue. In that article Pete digs more deeply into the similarities between the way film records light, and how your eye SEES light.
If you're not familiar with Cineon files, or working with film files in general, this is definitely the place to start.
Click on the image here for a 6 MB PDF download of that issue.
With that background under your belt, here's the whole story on how to work with Cineon files. |
The Cineon file format is the best way to digitize a 35mm film negative when both high fidelity and small file size are taken into consideration. The easiest way to think of a Cineon file is that it is the 35mm negative, only digital, and inverted (so that we can see something that vaguely looks like what was shot). But it is a hi-rez 10 bit facsimile of the 35mm negative.
Click image for larger
Why do Cineons look washed out?
The luminance values of a 10-bit LOG cineon file look strange, the darks are too light and the brights are not bright enough. This is because film captures a logarithmic record of the light intensity hitting the film.
A brief explanation of logarithmic
If you make a graph (see red line below) and you space the exponents evenly along that graph (not the number values themselves), the graph is said to be logarithmic. For example if you graph 10 to the power of 1, 10 to the power of 2, 10 to the power of 3, 10 to the power of 4 along the x axis of a graph, the 1,2,3,4 exponents are what are evenly spaced along the x axis (see bottom white numbers of graph), not the number values 10, 100, 1000, 10000, which are in totally different positions (mostly cramped over to the left of the graph as shown by the green line).
This graph is meant to demonstrate the principle of what is meant by LOG. Now let’s look at how this concept is relevant in the context of cineon files.
35mm film has a surface made up of microscopic silver-halide crystals. When light hits them, some of the individual crystals switch to dark. The rate at which crystals in the emulsion switch to dark decreases logarithmically as light increases. A photo of a 10% gray card will switch about 30% of the crystals to dark. A white t-shirt (which is 90% gray) will change about 70% of the crystals to dark. A bright glint on a car bumper 13 times brighter that white (1300% gray?) will switch 99% of the silver-halide crystals to dark.
Dude, It’s logarithmic.
So Cineons look washed out because 35mm film is very sensitive to darkish values and less and less sensitive to brighter values. In other words, 35mm film has an exponential (AKA logarithmic) sensitivity to light, which is why cineon files do.
Coincidentally humans perceive light contrast and detail logarithmically as well. We see more distinct levels of gray in the darker things we see (up to about middle gray), and as things get brighter we distinguish between them less. For example, in a darkish room at night we can still see things clearly and see contrast between objects even in the presence of a very small amount of light. If you are, however, driving on the highway on a sunny day and the clouds in the sky are too bright, you can’t see any detail in those clouds, but, by putting on sunglasses, you are effectively bringing the light values in the sky down into a range where you can perceive detail well.
There is an elegant similarity between how we perceive detail and how film captures light values, both of which could be represented by the green line in the previous diagram. Big budget movies are still shot on film despite the fact that we are well into the digital era, because the rate at which film captures light information is remarkably sympathetic to the logarithmic way that we perceive detail with our eyes. The aesthetic byproduct of this happy coincidence is that film is very pleasing to the eye.
"Ah just give me some 16 bit linear tiffs." Don't do it! If you can get cineons take ‘em. With 16 bit tiffs, you will lose all your super-whites and you need them to make photoreal comps. Have a look at the following jpegs of Marcie, which I shot with a so-so digital camera. In the first image, Marcie's hair is really blown out, in the second image I show the cartridge of a pen which, in the third image is all out of focus because I am holding it up really close to the camera. See how much detail was brought back into the hair even though the lighting didn't change at all between the first and third images?
Click image for larger
If I would have shot Marcie with a 35mm film camera and would have then had that image scanned by a cineon scanner, the resultant cineon file would contain absolutely all the over-bright detail that the pen cartridge revealed. If I try to, however, blur a pen in comp in front of the first blown out image, there is no way to make that look photo real. Since that image is a jpegs all the whites clip at 1. Cineons clip at 13.54
Click image for full size.
Tiffs, jpegs, targas etc... all can only contain values between 0 and 1. Technically speaking cineons also only contain values between 0 and 1.
Open any cineon in AE and view the values in the info palette. Set the readout to decimal by clicking on the little right pointing arrow at the top right of the info palette. You will not be able to find a value above 1.
But wait, if a cineon file is a faithful digital representation of a digital negative and a digital negative can hold values 13 times (or so) brighter that white, why doesn't the info palette show those super bright values? Well that’s because all the over-bright info is contained logarithmically in a 10 bit per channel file with 1024 levels of brightness per channel. The cineon file.
The slightly confusing world of the cineon After Effects workflow.
I've introduced you to Marcie, the star of Kodak's standard Cineon test file. If you'd like to play along with this part of the tutorial, you can you can download a Cineon file of my version of Marcie here. (Note: the file is 12 MB.)
Since it's a cineon file which can hold values above 1 (white), the project needs to be set to 32 bit (AKA float). The project working space should always be set to none.
Here is a side note: although there are ways of working with cineons in AE using ICC colour profiles I don’t recommend using them because I have found them to often cause a slight amount of loss.
Anyway where was I? Ah yes, apply the Cineon Converter effect from the Effects>Utility menu and you should have a beautiful linear High Dynamic range Marcie with which to comp. Mouse over the brights in her hair to confirm that you are in fact getting values above 1 in those highlights and...
...what the? Why the @*&$#*@#$ is it clipping at 1?! Let's retrace our steps. First check project settings, 32 bit, working space none, good.
Apply the cineon converter, which is indeed a 32-bit effect according to the little 32 next to it in the effects and presets window. So far that all seems correct, but where is my high dynamic range? Everything is still clipping at 1. Let's look at the cineon converter properties. White point, Gamma, hmmmm…, what's that last one, highlight rolloff, highlight rolloff = 20, hmmmm….
The highlight rolloff has to be changed to 0. The default for the cineon converter in AE clips values at 1 unless you set the highlight rolloff to 0. I assume this to be a legacy oversight. It is strange to have a 32-bit effect not have a default setting that is high dynamic range. When I finally figured out the highlight rolloff thing, I think I said something like "Well I’ll be." (I'm paraphrasing). Don't forget, change highlight rolloff to 0.
Click image for larger
So now that we have our cineon converter set up in AE we need to check our over-brights in the info palette.
Click image for full size
If we dim the image with the exposure control, the linearized cineon looks correct.
Click image for larger
But if a 16-bit tif made from the linearized cineon is dimmed down, the clipping becomes very noticeable.
Click image for larger
Even though, as seen below, the two images above were absolutely visually identical before they were dimmed.
Click image for larger
This is why HDR file formats like the cineon file format are so great for compositing. They aren’t a record of colour values, they’re a record of light intensity and we want the layers in our comps to interact with each other with real world light intensities in order for the interaction to look photo real. For sticklers there should of course be a gamma curve to linearize the ‘linear’ cineon comp, but that’s a can of worms that I’ll save that for another article.
Anyway, if we blur a foreground element in front of Marcie (in this case a simple blurred solid), the interaction of the cineon file with the blurred foreground element will be a pretty close facsimile of the way objects in the real world interact.
Click image for larger
And here below is the same comp with a tif instead of a cineon.
Click image for larger
So switch back to the linearized cineon file and now that our comp of Marcie with the blurred bar in front of her is complete, it is time to check out. Drag your comp onto the make new comp icon and call the comp that is created ‘output’. Apply the cineon converter to this comp making sure that all the numerical values are identical to the first cineon converter (expressions are good for this if you like), but change the setting from LOGtoLIN to LINtoLOG.
Click image for larger
But how can we be sure that the pixel values are unaffected in the area where the CG element doesn’t interact with the plate? Like this:
Click image for larger
Now you can render your cineon file choosing ‘Cineon Sequence’ from the Output Module>’Format’ pulldown.
The default setting for rendering cineons are as they should be for cineon 10 bit output. But here is a look anyway. The preset called ‘Full Range’ tells After Effects to render exactly what it sees on the screen to a 10 bit cineon file.
Hit render, et Voila!
|
