Technicolor 3way the correct way.

Discussion in 'Looks' started by David McLaren, Jan 28, 2013.

  1. David McLaren

    Message Count:
    405
    Location:
    New Zealand
    So ages ago there was a posting here about the technicolor 3way and I had a little play then actually nutted out a node tree that did the technical stuff of a technicolor film pass. This has been in my arsenal for the last 6 months and others are starting catch on. Here is a great tutorial of how to build it, done but some one else.

    http://www.mynahmedia.com/2013/01/3-strip-technicolor-look-in-davinci-resolve/
  2. Jason Myres Moderator

    Message Count:
    1,731
    Location:
    Los Angeles
  3. Andrew Taran

    Message Count:
    39
    Location:
    Vancouver, Canada
  4. Vladimir Bobenic

    Message Count:
    42
    Location:
    Bratislava, Slovakia
    great:D i was very closed...last node is the key..thank you
  5. Juan Salvo

    Message Count:
    1,361
    Location:
    New York City
    Sorry to be a party pooper on this one, but this isn't really a correct technicolor 3strip, as what's happening here has a lot of cross channel pollution, as Mike Most has pointed out in a thread about the same thing on creative cow.

    http://forums.creativecow.net/readpost/277/23060

    I have to add that just looking at it, it's clearly not a 3-strip look.
  6. David McLaren

    Message Count:
    405
    Location:
    New Zealand
    The difference with my build is I don't use the cyan magenta yellow build just straight RGB. Also instead of using the hue to change from negative to positive try using curves it's a "truer" transform from neg to positive. The result I usually get isn't quite so electric in the primary RGB. Also Juan it won't be exactly a 3strip look cause its not from film and has digital values but I use this as a basis for creating baseline to work from.
  7. David McLaren

    Message Count:
    405
    Location:
    New Zealand
    Here are some examples from a cooking show that I used this sort of build with. Actually I'll post them later iPad is being a bugger.

    image.jpg
  8. Juan Salvo

    Message Count:
    1,361
    Location:
    New York City
    I've gotten a very successful two strip look. Obviously I'm accounting for film vs digital here, even with that in mind, it's not really a three strip look. It lacks the characteristic tonality. I believe the process involved is inaccurate on several different levels. For one it's done entirely positively, it's using a parallel layer mixer, which seems an odd choice, and the 180deg hue shift should be unnecessary. Technicolor 3strip was additive CMY. A proper 3strip should add up in CMY.
  9. Juan Salvo

    Message Count:
    1,361
    Location:
    New York City
    This is definitely closer.
  10. David McLaren

    Message Count:
    405
    Location:
    New Zealand
    Ahh yes but in resolve when doing an additive with more than two inputs very strange things happen. That was my first attempt at trying to get a technicolor three strip. To be honest the above build is more akin to a 3ccd film path like that found in a spirit telecine. But the results are still interesting and valid in their own rights. I've not used this to actually try to do a filmic look from in ages. But definitely worth having in the power grades.
  11. John Burkhart

    Message Count:
    4
    Location:
    Singapore
    Hi Guys,

    This is my tutorital. Thanks for the interest, suggestions to improve it and constructive criticism is always welcome! I'm actually a little surprised that it got picked up around the web so quickly!

    This is my attempt at replicating an analog process digitally, and to learn more about the color theory involved. I actually got the idea after talking with the guys at Technicolor Asia, and was inspired by all the film infrastructure that was now sadly going to very little use, and reminiscing over film prints and analog technology with the engineers.

    I should probably point out that I did not claim this to be the "correct way" or the only way, but it's my way, :)

    This in and of itself won't make it look exactly like technicolor film of course, as suggested by another person it could be improved if the resolve allowed you to send picture information to specific RGB channels in a parallel node rather than mixing them all together. But then you start getting into Nuke/Shake territory, and starts to get away from Resolve's true purpose.

    I also need to add to the tutorial that it's not a "filter" that you just plop onto a grade, you still have to adjust the image yourself to get better "tonality" to use Juan's words. I'm thinking of adding a bit at the end about adding to the look by crushing the blacks a little bit, adding a little cyan to the shadows, increasing the contrast, but leaving the highlights nice and wide and maybe adding a little halation to them. But these are all subjective choices and will change with the footage. The tutorial is more designed to get you to about 80% or so of the final look, and you take yourself the rest of the way.

    Thanks!

    John Burkhart
  12. John Burkhart

    Message Count:
    4
    Location:
    Singapore
    Thanks
    Hi Juan,

    I decided to do it positively as the RGB strips off the camera were originally negatives of course, but they were then made into positive monochrome printing matirices before dying. My reasoning is that since no one shoots "negative" video, I could skip that step and start at the positive part of the process.

    I totally agree with you that the 180 hue shift should be unnecessary. In plotting it out in my head I was surprised I wasn't getting a positive color image, but an inverse one. It has something to do with the inner workings of the parallel node that I'm not privy to. But definitely the layer mixer was a non-starter in any of the modes.

    The next time I see Peter from Blackmagic I'll get his ideas on what's going on underneath (we both live in Singapore).

    Love the podcast by the way,

    John
  13. David McLaren

    Message Count:
    405
    Location:
    New Zealand
    Hi John and welcome to LGG. Nice build, did you do any tests to see if there were any difference between Cmy vs RGB other than just a tonal shift? Maybe it's time for me to dig back into this build and start experimenting. I remember being frustrated at being so close and yet not quite there.
  14. David McLaren

    Message Count:
    405
    Location:
    New Zealand
    Also john just wondering what's the technical reason for the red -cyan not blue. Why go to near opposites?
  15. John Burkhart

    Message Count:
    4
    Location:
    Singapore
    Hi David,

    Thanks for the advice, there is a slight color difference between the CMY and RGB, but it's really slight you can see it on the vectorscope, but I doubt you'd be able to tell on the screen. Yep, I feel like I'm not quite there myself. Let me poke around back in there and see if I can shake anything loose. Maybe it's just a few pieces of the puzzle are missing or need to be re-arranged. Maybe it simply can't be done "authentically" in resolve. I'm happy enough with the look itself, but enough of an obsessive tech nerd to want to get it exactly right too.

    Cyan's not a color you pick per-se, it's the color that's left when you remove all the red from each of the RGB channels. In the literature they also used Cyan for the tinting process too, so I felt I was on the right track.
  16. David McLaren

    Message Count:
    405
    Location:
    New Zealand
    Erm Juan go to the curves grab the arrow at the top of the luminance Chanel and drag to the bottom to do a neg to positive flip . Might save you some nodes too.
  17. Juan Salvo

    Message Count:
    1,361
    Location:
    New York City
    I prefer to use the "invert lut" but it wouldn't save me any nodes. This is based on the same formula used in "The Aviator"'s 3 strip process.

    You can do something in 1 or two nodes that emulates the effect, but the point here was doing a "correct" 3-strip... not just a 3-stripesque look.

    BTW, heads up on using that dragging curves to get inverted method, it's really hard to position at an exact negative, you usually end up over or under gaining. The "invert color" lut, is exact accurate and included, specifically requested it back after it was removed in the 9.0 update as I regularly use it for inverted effects. :)
    David McLaren likes this.
  18. David McLaren

    Message Count:
    405
    Location:
    New Zealand
    That would explain the trouble I had with slight polarisation with the darks until I tweaked the "invert lut"
  19. Just curious as to why John's method adds so much noise to the image. Could anyone explain?
    Juan, I love your three strip look! The "rubiness" of the red makes all the difference.
  20. Andrew Taran

    Message Count:
    39
    Location:
    Vancouver, Canada
    You all people are so smart! Hopefully in 20 years I could be as knowledgeable as some of you're. Great read!

Share This Page