Skip navigation
Currently Being Moderated

Good Workflow or Insane idea?

Sep 7, 2012 12:47 AM

Hi,

 

I have a workflow related question that I hope can be answered by those with a lot more experience than me....

 

The project we are busy with consists of a lot of post work for every shot we do (pretty much all of them). Here is a quick run down again:

 

1) We shoot miniature characters against a green screen at 50fps

2) Using the suggestion from Rick Gerard (found here: http://forums.adobe.com/message/4654521#4654521) I import the footage and render out a TIFF sequence as the first step. This sequence is then used as the base for all other work

 

From this point, there are many, many things that need to be done, and this brings me to my question - I need to:

 

1) Slow down the sequence more using Twixtor

2) Create a stabalized pre-comp for adding facial effects (mouth movements and voice synching)

3) Key the sequence for future compositing into background sets

4) Add subtle body movements if the scene requires it and we could not capture them during the shoot

5) Add scene specific special effects (be it an explosion or a laser fight....whatever!)

6) Create final composition for export to the editing program (probably Premier)

 

All of this detail, to come to my question - Does it make more sense to render out a TIFF sequence at each of the above steps, or should I try to stack as many of the effects in one single project/layer as I can. I am busy investigating the best workflow options for a project of this kind. I am concerend that too many intermidiate renders might cause quality loss, however, if i render out TIFF sequences, there should be no degradation right?

 

Anyway, if I can get any feedback on experience out there, I would appreciate it, and it will help point me in the direction of the workflow that would suit me best.

 

Thank guys - Have a great weekend

 

Pierre

 
Replies 1 2 Previous Next
  • Currently Being Moderated
    Sep 7, 2012 12:56 AM   in reply to Pierre Devereux

    You should do the keying, stabilization and slowdowns as one step, the rest as another. Since one is a technical thing and the other fall more into the creative realm, different workflows based on the footage would be necessary and different people work on it. It's usually better if one person covers the same aspects on all shots so at least if e.g. a green screen shot is messed up, it is consistent and another person can develop a fix that would work on all shots. That would also come in handy if you need to hand off stuff to otehr facilities or call in freelancers that each may have their own approach to such matetrs....

     

    Mylenium

     
    |
    Mark as:
  • Currently Being Moderated
    Sep 7, 2012 10:25 AM   in reply to Pierre Devereux

    1) We shoot miniature characters against a green screen at 50fps

     

    Unless those miniature characters are animated themselves via hydraulics or air or electric motors or something, why are you shooting at 50fps?  Why aren't you shooting stop-frame animation?

     

    The workflow would suggest that you're trying to do Wallace and Gromit - style animation, but sidestepping the effort it takes in planning to do it properly.  Am I incorrect?  Can you clarify what you're trying to accomplish in the end?

     
    |
    Mark as:
  • Currently Being Moderated
    Sep 7, 2012 11:26 AM   in reply to Pierre Devereux

    We have small pewter characters, lets say around two hand lenghts. They have joint skeletons and can move pretty realistically. The movements we want to incorporate range from minimal movement to crazy motion. We are trying something that we believe is pretty new (motion-wise) and without saying too much about how we are moving them yet, one of the movement types is done via large springs attached at certain places to the characters themselves. These springs are simply held, and the natural hand shakes travel into the character and give them a pretty nice "Natural" standing movement - no one stands truly still, we always move, even when standing still. - this is the feel we want to introduce to the characters.

     

    If i shoot at 50FPS (which is the maximum of the camera we have, this becomes half speed at 25FPS) and then slightly more with twixtor, the outcome is a very pleasing, extremely slight and barely noticeable "Life like" standing motion.

     

    Far be it from me to pry about trade secrets. 

     

    From your description, it would seen that a combination of stop-motion and in-camera animation would do the trick.  A rig to hold the springs back would let you control the too-fast movement caused by hands without having to resort to slo-mo... until you pop the springs, of course.

     
    |
    Mark as:
  • Currently Being Moderated
    Sep 7, 2012 1:50 PM   in reply to Pierre Devereux

    Tiff's are okay for intrermediate renders.  I'd use PNG's, but that's just me.

     
    |
    Mark as:
  • Currently Being Moderated
    Sep 8, 2012 8:34 AM   in reply to Pierre Devereux

    There are always exceptions, but as a general rule:  Key first!

     

    Your footage will always be at it's most prisitine when it has been untouched by other processes.  Once you start fooling with timing and stabilising, you potentially blur or merge pixels around your key edges, reducing the key quality. 

     

    Personally, and for the same reasons, I would probably stabilise in the next step.  Twixtor may change some pixels, potentially affecting stabilisation accuracy.  Leave Twixtor till after the stabilise, unless there's a pressing reason to do otherwise.

     
    |
    Mark as:
  • Currently Being Moderated
    Sep 12, 2012 7:20 AM   in reply to Pierre Devereux

    An image sequence is the same as a video. Changing when you convert from video to an image sequence won't effect stabilizing one bit. It also will not change the way Twixtor handles the footage at all.

     

    I'd get the footage stabilized first, then key, then slow it down to the final speed, then do the rest of your compositing. Motion blur from a really shaky camera may mean that you need to key first, then stabilize, then slow down.

     

    If your footage is highly compressed I'd add a preliminary step of converting the footage to a 10 bit (minimum) production coded before I did anything else, especially keying.

     
    |
    Mark as:
  • Currently Being Moderated
    Sep 12, 2012 7:22 AM   in reply to Pierre Devereux

    I'm not sure I understand the problem.  Surely you add the mouth in the final step, as it's animated and you can adjust the speed however you want it.

     

    I don't see why you cant:

     

    • Key

    • Stabilize

    • Re-time

    • Animate mouth

     

    or am I missing something?

     
    |
    Mark as:
  • Currently Being Moderated
    Sep 12, 2012 7:49 AM   in reply to Pierre Devereux

    You can key in the first comp.  You won't gain anything by keying in the second comp.  In the second comp, you can stabilize the footage & precomp, then apply Twixtor.  In the third comp, you ought to be able to accomplish everything else.  No fourth (final) comp necessary.

     
    |
    Mark as:
  • Currently Being Moderated
    Sep 12, 2012 8:02 AM   in reply to Pierre Devereux

    In my model, you use Twixtor in the second comp -- NOT the third.  So you're okay.

     
    |
    Mark as:
  • Currently Being Moderated
    Sep 12, 2012 8:10 AM   in reply to Dave LaRonde
    In first Comp:  Shoot, import and render out PNG

     

    Just make sure you up the color depth. Change your first project to 32 bit (float) and then choose an output option that gives you the option of more than millions of colors.

     

    png sequences give you the option of 16 bit files, but TGA and TIF sequences give you the option of 32bit (float) files. I'd go with the highest bit depth my system and my plug-ins can handle.

     

    Check the other plug-ins you may need to use in the production pipeline and make sure that they are all at least 16 bit compatible or learn how to use the HDR Compander.

     

    Here's a 32 bit tiff I made from a video that had about 5 stops of exposure added to blow it way out of the park. I brought it back into AE I added an adjustment layer and brought the levels back down to restore the image and even improve the color grading. The adjustment layer only covers part of the image so you can see the dramatic difference.

     

    Screen Shot 2012-09-12 at 8.00.08 AM.png

     
    |
    Mark as:
  • Currently Being Moderated
    Sep 13, 2012 7:27 AM   in reply to Pierre Devereux

    If i run Twixtor on the outcome, I end up with a 10 second comp at half speed - great, but the original stabalized precomp (where I need to add the mouth) is only half that length.

     

    Check the frame rates in your comps.  There's an obvious mismatch somewhere.

     
    |
    Mark as:
  • Currently Being Moderated
    Sep 13, 2012 7:59 AM   in reply to Dave LaRonde

    Actually misunderstanding about stabilising arises from the point that 'stabilising' Pierre needs and is talking about is not a piece of a technical part of the job, but a creative one.

     

    In the tutorial Mathias creates a stabilised pre-comp out of a character's mouth tracked in Mocha for only purpose - to apply Auto Lip-Sync script correctly. Therefore, there is no any frame rates mismatch or whatever. This is about understanding what is in the Mathias's tutorial.

     

    To sum up, after keying and slowing a footage down with Twixtor the technical part of the job finishes.

     
    |
    Mark as:
  • Currently Being Moderated
    Sep 13, 2012 1:16 PM   in reply to Pierre Devereux

    If it were me, I'd avoid rendering intermediate for the sake of rendering intermediate. Even though you definitely need one to track your character mouth in Mocha, you don't have to utilise it in AE project - you need just Mocha tracking data (hope, it's clear).

     

    Hence, perform some test. If your machine is beefy enough and can handle all the effects applied inside a master composition with several nesting levels without bogging your system down, purring animal sees no reason why render DI...

     
    |
    Mark as:
  • Currently Being Moderated
    Sep 17, 2012 4:50 AM   in reply to Pierre Devereux

    You obviously can't send a precomposition to Mocha, you have to render a footage out of that precomp to work with in Mocha. But my point was you don't have to reimport this footage in AE, you can apply Mocha tracking data (which you get by working on rendered footage) onto a precomposition directly...

     
    |
    Mark as:
  • Currently Being Moderated
    Sep 17, 2012 9:12 AM   in reply to Pierre Devereux

    If it were my project I would stabilize before anything else. The only time I would not do this is if I have to put the camera motion back in. In that case I'd do the slow mo processing, then stabilize, add your mouth, put the motion back, then key and composite.

     

    The key to a good workflow is to make sure that you are doing most of your processing with the frames that will be used in the final project. You want to try and get to the place you would be in your production if the camera work was perfect before you start doing anything else.

     
    |
    Mark as:
  • Currently Being Moderated
    Sep 17, 2012 9:36 AM   in reply to Rick Gerard

    Rick, there is no footage stabilisation in Pierre particular workflow. There is tracking a character mouth in Mocha and applying corner pin data.

     
    |
    Mark as:
  • Currently Being Moderated
    Sep 17, 2012 10:47 AM   in reply to Fuzzy Barsik

    See Step 2 item 2

    Step 2 – Stabalize and add Mouth

     

    Import TGA

    Use MochaImport to create stabilized precomp

    Use AE Auto Lip-synch for addition of mouth

    Precompose and apply compositing and effects required for shot.

     
    |
    Mark as:
  • Currently Being Moderated
    Sep 17, 2012 11:30 AM   in reply to Pierre Devereux

    Bookmark this Chris Meyer tutorial - it's going to help you solve some issues when it comes to stabilising some green screens (and understand why stabilising should be done prior to keying).

     
    |
    Mark as:
  • Currently Being Moderated
    Sep 26, 2012 3:12 AM   in reply to Pierre Devereux

    In relation to when to rely on Adobe Dynamic Link, and when - on a digital intermediate, have a look at this thread in PrPro Forum. Read up to the end and pay attention to Todd's comments.

     

    What format a broadcaster wants may depend on a broadcaster. E.g. here is BBC guidelines.

     

    Although Rick and other worthies with their really outstanding experience may answer about any question, do not hesitate using PrPro Forum, since sometimes you may need answers for PrPro specific questions.

     
    |
    Mark as:
1 2 Previous Next

More Like This

  • Retrieving data ...

Bookmarked By (0)

Answers + Points = Status

  • 10 points awarded for Correct Answers
  • 5 points awarded for Helpful Answers
  • 10,000+ points
  • 1,001-10,000 points
  • 501-1,000 points
  • 5-500 points