How to dynamically reframe for SD output?

[Elizabeth Lowrey]

In a discussion about the suitability of the RED camera for one type of work I do (dance performances), someone proposed capturing a static shot of the whole stage at 4K res (or perhaps at 2K to get deeper focus depth) and panning and cropping to "create" my camera moves and close/medium shots in post production, since my delivery format is SD with its measly 720 anamorphic pixel horizontal resolution.

I wonder how easy this would be to do and whether it would have to be done with the pan/crop tools of an NLE on the full sized REDCODE file or whether it could be done in REDCINE and output from there as 480P footage? Obviously it would require keyframing and pan/crop tools in REDCINE. Anyone know about this?

 

[donatello b]

take a look at Brooks FAQ .. he has some good examples of 4k frame then pulling 2k ( about 1/2 way down page - the upper page are film examples) ...

http://www.reduser.net/forum/showthread.php?t=1487&page=2

 

[Elizabeth Lowrey]

Yes, thanks, donatello. I had read his FAQ and frequently refer back to it, as it is an excellent resource.

My question is not about depth of field characteristics and how windowing versus rescaling and different lens choices affect that, though. I'm talking about the actual nuts and bolts of dynamically rescaling a static 4K shot so that you can cull essentially multiple SD "cameras" and "camera moves" from a single high resolution shot. I do this all the time with high res stills in Vegas (and I'm sure Avid, Premiere Pro, Final Cut, etc, all permit the same thing). But I'm just curious about whether this pan/scan technique is practical with moving 4K REDCODE RAW footage or whether my computer (Core 2 Duo 1.86 GHz, 2GB RAM) would choke to death even trying it.

Perhaps the preferred workflow would be to render out multiple 480P camera "angles" from the master 4K shot with REDCINE and then edit those as a regular old SD multicam project. I can't really imagine that REDCINE is built to do a long form, meticulously key-framed pan/scan of 4K footage, so I'm assuming doing the edit within REDCINE is not possible or practical. But I don't know that for sure. Was just curious.

 

[Hrvoje Simic]

In a discussion about the suitability of the RED camera for one type of work I do (dance performances), someone proposed capturing a static shot of the whole stage at 4K res (or perhaps at 2K to get deeper focus depth) and panning and cropping to "create" my camera moves and close/medium shots in post production, since my delivery format is SD with its measly 720 anamorphic pixel horizontal resolution.

Maybe you should take into consideration that cropping won't give you DOF manipulation and shooting a whole performance from one angle may become boring to a viewer. Also, with one angle control over composition will be very limited and you loose shooting creativity. One shooting angle may be good for one subject in the scene, but very bad for the other. The importance of this factors depends on what type of project it is, but if you want it to be dynamic and interesting to watch, you will need many different angles IMHO.

 

[Steve Freebairn]

While it would be possible, it would not look right. take an image from a dslr and then take it into Ae or motion and do some pans and scans on it (partly because of paralax). It looks different than if you had repointed the camera. Besides it looking weird, it would take a really long time.
You could do it in AE though or motion/shake. You'll be able to test it out this weekend when footage and redcine are released.

 

[Hrvoje Simic]

My question is not about depth of field characteristics and how windowing versus rescaling and different lens choices affect that, though. I'm talking about the actual nuts and bolts of dynamically rescaling a static 4K shot so that you can cull essentially multiple SD "cameras" and "camera moves" from a single high resolution shot.

Oh, I see.

 

[Elizabeth Lowrey]

In casually browsing the forum, looks like Rob Lohman provided the REDCINE part of the answer to my question:

It will do reframing, blow-ups & cropping. However, a push/pull indicates animation with keyframing and such and that is *not* in REDCINE.

REDCINE is the RAW conversion tool to get your footage in the ballpark (look wise, size wise and so on). It's not a software package that'll do your whole post workflow.

 

[Blair S. Paulsen]

Just one alternative

Just one alternative

AE is the right beast but the renders will take a looong time, especially on an older computer - of course you can let it render at night. The single angle issue was a given in your initial post so that hasn't changed. The pan & scan in post will look different than actually moving the lens, absolutely. You should take 20 seconds of footage and play with it to see if you like or hate the results before you shoot an entire performance.

 

[Elizabeth Lowrey]

AE is the right beast but the renders will take a looong time, especially on an older computer - of course you can let it render at night. The single angle issue was a given in your initial post so that hasn't changed. The pan & scan in post will look different than actually moving the lens, absolutely. You should take 20 seconds of footage and play with it to see if you like or hate the results before you shoot an entire performance.

Actually Vegas does outstanding pan and scan moves right from within the NLE with very intuitive, quick keyframing tools and excellent accuracy. So if I do this at all, it would have to be from within Vegas (or whatever NLE I wind up using with RED ONE). Anything else (I use Boris RED for most 3D compositing) would be far too cumbersome from a workflow perspective.

I realize that the parallax difference will not exactly mimic moving the camera. But then that issue is no different from the one involved in panning and scanning of stills, and that's been a major technique in documentary/promo work for the past umpteen years. The eyes and perceptions of viewers have been trained, I think, to look at this inconsistency as "normal".

Nor is this issue really any different from 3D compositing of green screen/roto footage without precise motion tracking. The parallax is always changing. But if you use common sense, you can keep it from being very noticeable.

I just composited a piece a few weeks ago in Boris RED utilizing stage footage keyed with multiple linear color keys with a girl doing a leap. Put her in a 3D environment with her standing on one of the letters of the titles and jumping up to the adjacent letter as the camera panned left to right and up. It wasn't perfect, but with lights and shadows to help fool the eye, it still looked pretty neat and convincing enough for a surrealistic composite.

I agree that trying to cull multiple camera angles for an entire performance is a different beast, so I will definitely be doing extensive tests before I decide whether it's a viable and aesthetically pleasing strategy.

But there are already several things about it that have potential appeal for me:

(1) It would allow me to keep all footage perfectly consistent in imaging characteristics since it would all be from the same camera. It's always a bit disappointing to be mixing my freelance camera guy's JCV DV5000 footage with my Sony DSR300 and Panasonic DVX100 footage. They are all different and it's impossible to make them look exactly the same (not a problem with documentary or even narrative footage where environments are always changing from shot to shot but more bothersome with multi-cam performance footage). Further more, I can't even utilize the DVX in progressive mode because I'm always mixing it with interlaced footage.

(2) I would never, ever have to worry about not catching some important entrance or exit in close enough detail since I could leave the RED full stage and crop it to the area of interest for that particular action.

The proof is in the pudding, though. I won't know any of this for sure until I get my RED and try it out. Unfortunately that's still many months away.:waaa: