Home Page › forums › Autodesk/Discreet › Flame and Smoke › Motion control cock up
- This topic has 8 replies, 8 voices, and was last updated 11 years, 2 months ago by Taneli Santamäki.
-
AuthorPosts
-
August 2, 2009 at 12:16 pm #202988Cover PointParticipant
Hello Folks,
I am freelancing on a job involving a frozen time sequence. It was shot on Red mounted on a motion control rig. It´s a complex camera move of about 2500 frames around a living room with actors and props supported on various rigs which need to be rotoed out. The problem is that after the hero pass a number of clean plate passes were made BUT the camera has moved on the mount so the clean plates to do not match the hero pass or for that matter each other.
Has anyone come up against this sort of thing before? My first thoughts were to use camera tracking data from the her pass and and place whatever clean elements I need in relation to this, however there is significant perspective change to take into account.
Any experience folks have had with this kind of thing would be greatly appreciated.
Cheers
Bill Shearman
August 2, 2009 at 2:35 pm #218102AnonymousInactivehow different are they? I’ll assume its different enough so that small patches are noticeable.
Worst case scenarios is, roto everything, and put all of that into your clean plate.
Or, how about lots of larger softer patches instead of smaller, precise patches, which may give away the discrepancy in the plates?
good luck. wear a helmet. sounds nasty.
randy
August 3, 2009 at 6:00 pm #218100Martin FurnessParticipantGot to love it!!! I deal with this all the time. Just amazes me how the camera somehow mysteriously moves, especially with a shot of that magnitude! Is the camera “that” far off on the plates? Can the plates be manipulated at all to line up? Its a shame when you attempt to do your best work and ignorance restricts you! and it ends up being “it is what it is” Without the elements in front of me its tough to consider, especially Monday morning. LOL Good luck, im sure you will find some answers here.
August 3, 2009 at 7:46 pm #218101Piotr KolusParticipantYou can try the Furnace Align spark, it has saved me more than once in the past in situations like this. You will still need to mask out (roughly) actors and other elements that are not in the clean pass and spend some time tweaking parameters, but it does work.
August 4, 2009 at 1:30 pm #218099bnwParticipantHm. AFAIK Red can’t sync its shutter to a moco rig. So I would imagine the lack of sync between passes is the problem rather than the camera moving… see http://www.reduser.net/forum/showthread.php?t=17002
If that is the case, you may have some luck using a decent retimer to shift the plates in time by less than a frame so they line up.
Or, if you have the latest Furnace on a Flame, the F_Correlate plugin tries to do exactly that for you… I’m not 100% sure whether it will interpolate between frames, which is what you want, but it’s worth a try…
August 5, 2009 at 9:51 am #218104bernardo sodreParticipantHi All,
Thank so much for all the replies. Loops has nailed the source of the problem…lack of synch between the camera and the rig. To top it all the passes display random, high frequency instability!
Anyway I am using Furnace Steadiness, then Align to get things in place. I cam tracked the stabilised hero pass in Syntheyes and am using this data in the keyer camera to avoid too many gmask keyframes. So far so good. Its a long slow process, I am have to work at 3K because the director might want to do some reframing(!) later.
Once again thanks to you all for the input.
Cheers
Bill
August 31, 2009 at 2:03 pm #218106Taneli SantamäkiParticipantThere is a trick, it involves virtual projectors.
If the two takes are not to far of from each other you could make a simple model of the set in 3D and project your footage onto that using a 3D track of the faulty camera. Then you use the camera track form the “correct camera” and use a virtual camera to “reshoot” the material.Look at:
http://www.syndicate.se/Default.aspx?Id=150
There most of the floors and other textures are remapped using this technique.
@Cover Point 28532 wrote:
Hello Folks,
I am freelancing on a job involving a frozen time sequence. It was shot on Red mounted on a motion control rig. It´s a complex camera move of about 2500 frames around a living room with actors and props supported on various rigs which need to be rotoed out. The problem is that after the hero pass a number of clean plate passes were made BUT the camera has moved on the mount so the clean plates to do not match the hero pass or for that matter each other.
Has anyone come up against this sort of thing before? My first thoughts were to use camera tracking data from the her pass and and place whatever clean elements I need in relation to this, however there is significant perspective change to take into account.
Any experience folks have had with this kind of thing would be greatly appreciated.
Cheers
Bill Shearman
September 1, 2009 at 1:59 pm #218105bernardo sodreParticipantHi Boombastik,
This was indeed how we ended up solving some of the problems of generating a clean plate and worked really well. Careful and accurate camera tracking data is a must.
Thanks for you reply
Cheers
September 2, 2009 at 4:07 am #218103Ruslan BorysovParticipantGiven the normal 24fps that most motion control runs at, it’s surprising this wasn’t flagged on set by the tech people there. A non-synced set up could give a disparity of a few feet on a decently fast move!
Still, it goes to show that, although film has been around for 100 years and is still basically the same technology, they have really ironed out the kinks. An all digital camera should do better at motion control but nobody has quite got around to getting the Red onboard yet, I guess!
-
AuthorPosts
- You must be logged in to reply to this topic.
