In comparing jpeg #4 (original) to jpeg #6 (technicolor) there is a problem. In jpeg #4 there are blown out highlights in the clouds. In #6 those same areas have details. This leads me to believe that there is something wrong (syspect?) with the information in jpeg #4.

I think one could come very close to matching the "look" using Color Curves and the Color Corrector, but due to the limitations of video, it'll never really "match" per se.

For example, the dried pine needles along the curve in the road... I cannot get the "red" into that area using the Color Corrector alone. Whereas if one were able to "seperate" the channels as show, each channel could be tweaked accordingly.

The problem for me is that I am not "technically minded" enough to do this. I could spend the next year or two playing around and I may stumble onto something, but I seriously doubt it.

The problem with the normal use of color correction is that it affects the whole frame. Color curves have a different problem, you can alter the gain of each channel however there's little point in adding gain if there's nothing in that channel. Worse still the way DV samples color means you might have very low res in the channel you're trying to work with.

However all is not lost, the luminance channel has plenty of resolution as it's got the best sampling. One can extract the luminance channel using the 100% B&W FX. Now take that and using say the Secondary CC make it the color you want and mask that into the video. Certainly this will not give you accurate color rendition and it can be painfully slow creating masks but also rendering back to DV you could loose at least some of what you gain.

Ignoring the Technicolor challenge, to make the pine needles red.
Create a new track by rendering the original with the 100% B&W FX. Drop that on an upper track, apply SCC to make it red and black then add a bezier (with feather) to cutout all but the pine needles, set composite mode to Add. Bingo, red pine needles. Adjust Composite level to control how red, adjust SCC to contol shade of red.

Sad times; I have missed the entire point of my own exercise.

farss, you seem Vegas-brilliant, I would like to understand how you are doing some of these things that you describe, like adding a bezier (with feather). I've never seen such an option/tool in Vegas (I'm running 6.0d). I've also yet to find the MinMix FX.

Sorry for falling behind, but I am a very fast learner and I would like to educate myself through this process as well.

This might be useful:

Aviator technicolour process...

Here's a follow-up to that one:

INSTRUCTION:from DFX on VFXTalk thread

Jonathan, after some additional think, and I bounced this off Grazie, I'm not sure the end result can be achieved in Vegas. Consider this (albeit overly simplistic)...

The video camera has three chips: red, green and blue. The camera is, basically, doing what the Technicolor process did/does. The camera then merges those three seperate images (red, green, blue) into one, giving the final, full-color image.

Even if we were able to accurately break that image back down into its three basic components, and bring them back together, it stands to reason that the end result would be no better than the original image was to begin with. Does that make any sense?

Having said that, I'm not saying it can't be done. I am saying I don't see how it would be done by me using Vegas alone (minus all the "work arounds" that have been suggested).

Jay

Breaking the image into three colors (red, green, blue) and then reassembling them is just making a complete circle back to where you started. However, what appears to happen in the aviator-fx video is that certain color mattes are applied to certain color channels which result in an emulated technicolor feel, supposedly. If this cannot be done in Vegas, then I say 'so be it', however, I'm all about the experimentation & discussion process (and I've really liked how it has gone so far). We can all learn a lot from this, regardless if the goal can or cannot be achieved.

Perhaps the goal (to simulate the technicolor feel in digital video) can be better emulated with plugins like Magic Bullet or pre-existing effect chains. In that case, I would love to have some discussion on which chains we think could better produce a technicolor feel. Meanwhile, I'm trying to experiment with the method demonstrated by the Aviator-fx team.

So, to summarize. This process is not just about breaking the colors and reapplying them, but it's about what happens to those colors after they are broken, but before they are reapplied.

... but it's about what happens to those colors after they are broken, but before they are reapplied.

And I think that information is the piece of the puzzle that is missing.

Yes, as I said before (and perhaps I am all wet on this), but this is really all about how to create an fX that will emulate the Technicolor "look". I think the whole separation into separate plates and then reassembling those plates is a little misleading because I don't think you necessarily have to travel that path to get to the desired destination.

If you click on the links I supplied in my previous post, I think you will get at least some of the missing information in the puzzle, or at least one person's attempt to supply the "formulas" for the mattes.

Near as I can tell from having looked at this, when you break apart what is being done, it looks like the relative strength of each primary color channel are being changed. This obviously involves more than just increasing or decreasing each channel, but also changing the "S" curve for each channel.

I have color-corrected over 60,000 still photo scans in the past fourteen months, and while I don't pretend to have totally figured out everything about color correction (quite the opposite, actually), I've created almost exactly the effect shown in the examples by changing each channel's S curves. To do this in Vegas, I would create a curve for each color channel using the Color Curves fX. Since the whole thing is more art than science, and is going to be different for each scene and each exposure, I would find something that works as a starting point, and then go from there. Attempting to find actual formulas based on film emulsion masking, or attempting to reverse engineer based on the transfer characteristics of each of the film stocks used for each of the three strips in the 3-strip process, probably won't get you anywhere.

What might be useful, perhaps, would be to somehow find a few still image captures of technicolor, and then those same exact frames from a non-technicolor print. I think one might find these on the "extras" section of various DVDs that have been created from restored versions of movies originally shot in Technicolor. The best example that I know of is "Vertigo."

More interesting reading, as always, on Wikipedia:

Technicolor

Just thinking out loud here.
Me thinks it wouldn't be that hard to build a Technicolor video camera. Take the signal from each CCD to it's own VCR and then matrix it back in post. Probably quite useless in any practical sense, would help sell a lot of tape though.

That aside I didn't think the purpose here was to produce the same result as the Technicolor process, I doubt if even what was done for The Aviator produced a result 100% the same as the original. Instead I thought this was a learning exercise and for that I think it's been a success.
Jonathan has realised that Vegas has way more under the hood than it's simple looking GUI reveals. I'd suggest he spends some time doing a quick read of the Vegas manual and perhaps invest in some training materal.

Reading through the Wiki only adds to my belief that the Technicolor look was more than just color correction. Several things were done to the image that must influence the final look of the image. I noticed that this was a kind of die transfer process during printing so there must have been some bleed of colors. There was also issues with registration. I note also that a monochrome overlay was applied to mask the registration issue. This could be achieved by applying a small amount of GB to the separated strips before recombining and also compositing in a B&W track from the original full color track.

Bob.
Bob.

As I explained in my previous message... the math behind the Technicolor process is not so straight-forward. It also depends on whether you shot on modern film or video.

In Aviator, they shot on modern film. Breaking apart the *original* R, G, and B channels is not so easy, since the red in the image affects the green and blue.
Similarly, the green affects the blue. This is the "crosstalk" described on the Aviator site.
In video cameras, there is no crosstalk of this kind. I believe the Technicolor process has this crosstalk, although it's a different kind.

Example if this makes things any clearer:
Think of the red layer as sort of a ND filter. If the imagery were a gradient, the red layer in modern film would act like a ND filter to the green and blue layers. If there was no red in the scene, the red layer would block less green or blue light.
What happens in the red layer of the film affects what happens in the green and blue.

**Modern film doesn't exactly have red, green and blue layers (and some films have four layers). But it's easier to think of film as if it has R, G, and B layers.

2- There will be metamerism happening in modern film stocks, so it won't be the same as recording Technicolor. This is analagous to the difference in colors between a Sony F900 and a Varicam. It's a subtle difference and we usually don't care.

3- The 2-Strip Technicolor look (not 3-strip):

The following method will get the gist of the 2-Strip Technicolor look, although the look of the saturation will be wrong.

Add the channel blend filter.
R = 1 * R
G = 0.5 * R + 0.5 * G + 0.5 * B
B = 1 * G + 1 * B

This knocks out the original green information, emulating the look of recording with only two color layers.

Then add the color corrector.
Bump Saturation to taste.
Move the high slider around to make flesh tones look correct. i.e. 2 -strip added yellow to make flesh tones look correct... this does something similar (but not the same).
You could also move the mids slightly.

Then add the color curves, and use a s-shaped curve to somewhat mimic film's transfer curve. I believe the difference is that this is only a 1-dimensional transform, whereas with film it's sort of 3-dimensional (because the layers interact with one another).

There has to be a way and it's probably fairly simple. I think part of it is the technicolor which, according to the wiki, is actually a CMYK printing process. That would naturally introduce a color shift due to the 'complementary' colors not being an exact complement.

I thought I was getting close by doing an initial split with a magenta filter and then further splitting that matte into the blue and red components. Then the mattes were adjusted but the contrast turned out to be wrong in the final result. The colors looked pretty good though but could probably have been done with color curves

There has to be a way. Of course, I never say die...

Very interesting thread. Haven't had time to experiment yet.

Interesting about CMYK vs. RGB. Could this be about the old Light color "wheel" vs. Paint color wheel differences?

Since Technicolor was a light-based color process, and most of us are in the "paint" color world with our PCs, shouldn't we be isolating the Cyan, Magenta and Yellow colors (plus black-K) instead of the Red/Green/Blue?

Just an old projectionist interested in the effect...

I wanted to get somewhere along with getting colour mixing of layers. So here goes . . .

Track 1 - SAMPLE TEXT Blue Text

Track 2 - SAMPLE TEXT Green Text

Make T1 Parent to T2 and apply "Difference" to T1:

Result = Turquoise

Repeat and alter colours:
T1 = Red ( Red at the yellow end - yeah?)
T2 = Blue

Result = Pink

Just repeated with successful results using SOLID Sony Colours

Questions:

Q1: Have I been successful in mixing colours?

Q2: Is this a clue into achieving the technicolour task?

The trick is to change compositing mode to ADD.

For example, put a R,G,B circle onto each of three tracks and overlap them. Should get a rather classic demo of additive color at work but you don't. Change composite mode to ADD on all tracks and bingo.

Bob.

Base on those numbers, you can create the matts using Channel Blend and Mask Generator - for Green Matt,
Channel Blend G= -0.5*R + 1*G + -0.5*B, A=1*A all the rest 0.
Mask Generator - Green Channel, invert,Low in - at 0.7 to 0.8 (depending on strength of effect).

put Green Matt on first track, Blue Matt on second Track and RED only on third, make the first and second tracks composition as Multiply - this gives you the new RED channel. Place all 3 tracks under a master track.

Do the same for Green and Blue and use composite mode of ADD between them.

In order to fine tune, use different low level in the masks.

BTW - According to articles, the Aviator FX department create a Look Up Table (LUT) using Photo Shop. They created many LUTs for 2 strip and 3 strip with different strengths. And they use the LUTs and not the complicated process.

So, Bob and Udi, has my suggestion pushed/eased/seduced the process further forward or not?

The CMY is what I got early on in the procress when mixing two of the RGB layers as indicated in the video. My problem was I could not get them "composited" back together for some reason.

to the wiki, is actually a CMYK printing process.

The Wiki article does link to CMYK, but the process they describe does not involve the K (black) undercolor process that is part of printing images on paper. Somehow I doubt that this was really part of the technicolor process (the sentence in the Wiki actually only mentions CMY, and then -- carelessly, I think -- mentions the relationship to CMYK). Therefore, I don't think you need to be emulating undercolor removal or the addition of a black layer in order to create the desired Technicolor fX.

Wow again, look at our think-tank go! Have any of us achieved any effects worth mentioning and if so why don't we start sharing images? There are more than a few free image hosting websites available, and last I checked, these forums allow for HTML code, so you can post your img href tags right in here. Do you think that would be a good idea?

Also, am I alone in wishing that the color curve points could be adjusted numerically (right click, goto something and get/edit the two-dimensional matrix points)? If we're going to start sharing color curves it's difficult to mimic those when we're limited to analog adjustments. Or, can Vegas already do this?

I see some great ideas (more than one method) developing, and who knows, we may be on to what will later be a very popular Vegas effect.

Sitting at my work computer (no Vegas) I set out to try it with Photoshop and achieved the same results. So I guess it is possible to create an Photoshop action and export the frames but I have no idea how long that would take.
Here's what I did:
1. Convert the image to Lab mode (colors are different channel than B&W detail)
2. Make the backgound a layer and create 2 duplicate layers naming them Red, Green, and Blue.
3. For each layer use a Hue/Saturation adjustment to adjust and isolate the colors you want and leave everything else desaturated.
4. For the top two layers (it doesn't mater what they happen to be), delete the Lightness channel which holds all the detail. You are left with a mostly grey image with your target color showing.
5. A Color Range selelction set to that grey can delete everything on the layer except the target color. This needs to be done to those top two layers. The bottom layer will retain the B&W detail as well as it's target color.
6. Flatten the image.
7. Convert back to RGB mode.

This is the easiest method I could produce with the Beach photo that worked only with the colors and didn't require further adjustment with overall contrast or brightness. Again, creating an action and exporting frames might very well take months to complete:)

Ben

To print the film, each colored strip had a print struck from it onto a light sensitive piece of gelatin film. When processed, "dark" portions of the film hardened, and light areas were washed away. The gelatin film strip was then soaked with a dye complementary to the color recorded by the film: cyan for red, magenta for green, and yellow for blue

Granted it is not printing in the typical sense but this is where I got the notion, perhaps incorrectly, that it would be closer to a CMYK subtractive process rather than a RGB additive. Maybe they composited the individual strips after the dyes were applied?

Okay, I got the RGB mattes made, they look "similar" to those little, itty, bitty ones in the video, but I can't get them to composite in 2s layers (the blue and green to make the new red, for example) to make the "new" RGB layers respectively.

And I think this is pivitol in the process. Because if you look at the original layers as compared to the new layers, there is a significant difference between the two! Look at the original red layer. The shadow of the tree in the grass is very distinct. Whereas in the new red layer, the shadow is non-existent, or nearly so based on the size and lack of detail.