Jump to content


Photo

Motion Capturing in "Ted"


  • Please log in to reply
8 replies to this topic

#1 Benjamin Davis

Benjamin Davis
  • Basic Members
  • PipPip
  • 20 posts
  • Student

Posted 17 July 2012 - 03:33 PM

I haven't actually had the chance to see "Ted" yet, but I was wondering if anyone here has worked on a similar project, with choreographing lighting and camera work around something that's not in the scene until post-production. Gathering as much information on the film through behind the scenes videos and photos, information on the visual effects studio's websites who worked on the film, and snippets of information from the cinematographer and visual effects artists from various film websites and whatnot, I've begun to seriously appreciate the amount of work the cast and crew has to go through to successfully create a realistic and good looking shot based around a digital teddy bear! I've never worked on such a project, and my post-production experience really only goes as far as color grading and editing, so I've not been able to really understand the efforts behind pulling off such a film.

So back to my original question! Has anyone here worked on such a project? I can only assume it's similar to shooting a traditional scene, except with a slightly different set of objectives and tactics to actually shooting. What's the process of shooting like, and what's the preparation for shooting something like this entail?
  • 0

#2 David Gregg

David Gregg
  • Basic Members
  • PipPip
  • 21 posts
  • Other

Posted 02 August 2012 - 08:48 PM

With the advent of 3d motion tracking a dozen or so years ago, it really became practical to do a lot of shots with realistic animation inserted and interacting with the scene. I have to presume that 3d motion tracking is the primary method (that and a lot of great animation).

I've worked with Real Viz's Match Move (now Maya's Match Mover). Match Mover and related software products are amazingly accurate for calculating and deriving the camera, its lens and its motion path. If done right, think 1/4 pixel accuracy. The key is that there has to be enough depth information in tracking points for the software to figure out where the camera is. That information usually comes from orbital, trucking or dolly shots since a single stream track has to have enough parallax to figure out the 3d camera and its path. The tracking software is amazing and works even on something like an unregistered film camera. (I was once playing with some old stock film footage on video that had the fields reversed and some screwed up 3-2 pull-down, but the darn thing still tracked it and derived the camera!)

The high-end systems allow for stills or other secondary shots taken from different angles to be used to provide depth information. Amazingly, most of the software systems can use most any camera's alternate angle to add depth data for tracking. A digital still camera can be mixed with an HD video or film camera for example and it still works.

About the only thing that won't work is a simple pan from a tripod head or a lens zoom. There is no parallax in a pan or simple zoom so there is no depth data (unless you have "helper" shots of the same scene from other angles).

My link
  • 0

#3 Chris Millar

Chris Millar
  • Basic Members
  • PipPipPipPip
  • 1642 posts
  • Other

Posted 03 August 2012 - 12:12 AM

While I agree pans are an exceptional case in match moving I'd hesitate to say the there is no parallax in a typical pan - more often than not the cameras nodal point (I forget the correct term) is not coincidental with the axis of rotation of the pan head - not unlike our eyes being offset forward from our neck resulting in a small hut appreciable parallax.

;)
  • 0

#4 Chris Millar

Chris Millar
  • Basic Members
  • PipPipPipPip
  • 1642 posts
  • Other

Posted 03 August 2012 - 12:13 AM

Gawd, I really hate 'typing' on an iPhone :rolleyes:
  • 0

#5 David Gregg

David Gregg
  • Basic Members
  • PipPip
  • 21 posts
  • Other

Posted 08 August 2012 - 07:18 PM

While I agree pans are an exceptional case in match moving I'd hesitate to say the there is no parallax in a typical pan - more often than not the cameras nodal point (I forget the correct term) is not coincidental with the axis of rotation of the pan head - not unlike our eyes being offset forward from our neck resulting in a small hut appreciable parallax.

;)


In practical terms, there isn't enough parallax in simple pivotal pan to my experience.

Most of the 3d tracking software does have the ability to add secondary views from other angles as "helper" files to give it depth data. I only known Maya's (RealViz) Matchmover, but it is happy with most any "helper" file even if it comes from an entirely different camera and lens. Pretty amazing stuff. In any event, I wouldn't recommend counting on a pan for 3d motion tracking. Dolly shots are great though!

I wouldn't count on the auto-trackers either.


My link
  • 0

#6 Chris Millar

Chris Millar
  • Basic Members
  • PipPipPipPip
  • 1642 posts
  • Other

Posted 08 August 2012 - 08:16 PM

In practical terms, there isn't enough parallax in simple pivotal pan to my experience.

Most of the 3d tracking software does have the ability to add secondary views from other angles as "helper" files to give it depth data. I only known Maya's (RealViz) Matchmover, but it is happy with most any "helper" file even if it comes from an entirely different camera and lens. Pretty amazing stuff. In any event, I wouldn't recommend counting on a pan for 3d motion tracking. Dolly shots are great though!

I wouldn't count on the auto-trackers either.


My link


Best case (or worst depending on your perspective, (excuse the pun)) for tracking/parallax is a wide angle lens that is mounted say on an ENG style camera where the centre of gravity is aft of the lens nodal point/points - in fact even a shoulder mount is offset on the LR axis itself ... ;)

With the offset you can even make a cheat/quasi-3D by using temporal offset as opposed to the traditional spatial offset - also easily achieved with dolly moves where the camera direction is perpendicular to the motion (which many bullet time effects virtually equate to also).

This stuff fascinates me :)
  • 0

#7 David Gregg

David Gregg
  • Basic Members
  • PipPip
  • 21 posts
  • Other

Posted 14 August 2012 - 01:37 AM

Best case (or worst depending on your perspective, (excuse the pun)) for tracking/parallax is a wide angle lens that is mounted say on an ENG style camera where the centre of gravity is aft of the lens nodal point/points - in fact even a shoulder mount is offset on the LR axis itself ... ;)

With the offset you can even make a cheat/quasi-3D by using temporal offset as opposed to the traditional spatial offset - also easily achieved with dolly moves where the camera direction is perpendicular to the motion (which many bullet time effects virtually equate to also).

This stuff fascinates me :)


If I was putting money or a production schedule on the line, I'd design any post motion tracking show with lots of parallax. Otherwise, things could get complicated and maybe even painful. Been there. Done that.

Some fun with Match Mover
  • 0

#8 Chris Millar

Chris Millar
  • Basic Members
  • PipPipPipPip
  • 1642 posts
  • Other

Posted 14 August 2012 - 02:44 AM

I'm looking at it from the perspective of a designer of such systems as opposed to a user - it's not the best tactic to say 'been there, done that' in my case ;)
  • 0

#9 David Gregg

David Gregg
  • Basic Members
  • PipPip
  • 21 posts
  • Other

Posted 24 August 2012 - 10:34 PM

I'm looking at it from the perspective of a designer of such systems as opposed to a user - it's not the best tactic to say 'been there, done that' in my case ;)



Roger that. I find the software to be amazing. Having painfully done more wild tracks without the benefit of such a dead-on lock down, my hats off to the developers! I just wish the sales staffs wouldn't push the "auto tracker" until it learns the difference between good corners, intersecting planes and reflections. All it takes is one bad track point to mess up a perfectly good week.
my trip to mars
  • 0


Willys Widgets

Glidecam

Paralinx LLC

Media Blackout - Custom Cables and AKS

rebotnix Technologies

Ritter Battery

Opal

CineLab

The Slider

Gamma Ray Digital Inc

FJS International, LLC

Visual Products

Abel Cine

Aerial Filmworks

Technodolly

Rig Wheels Passport

Broadcast Solutions Inc

Wooden Camera

CineTape

Tai Audio

Metropolis Post

Broadcast Solutions Inc

Tai Audio

Wooden Camera

Aerial Filmworks

Visual Products

CineLab

Technodolly

Abel Cine

The Slider

FJS International, LLC

Media Blackout - Custom Cables and AKS

Rig Wheels Passport

Metropolis Post

CineTape

Paralinx LLC

Ritter Battery

Glidecam

Gamma Ray Digital Inc

rebotnix Technologies

Opal

Willys Widgets