Sub-frame clip positioning in CS5?

[Tim Kolb]

Hold down the ALT key while sliding the audio part of the clip when set to audio units.

TimK,
Director, Consultant
Kolb Productions,

[Alex Zvyk]

Well, I’m not really trying to slip audio out of sync with its respective video clip.

What I’m really trying to do is slip the video/audio clip in its entirety against another clip from the 2-cam shoot, so audio marks of both clips would match with better than 1 video frame accuracy on the timeline.

Come think of it, this is probably impossible per-se, because 1 video frame is the smallest unit the video is quantized by.

Maybe artificially increasing timeline’s framerate, then, will help solving this?

For instance… my videos are 30p on a 30p timeline. What if I change timeline to 60p. Or 120p. Then it would appear that I should be able to better adjust the positioning of clips on different layers, matching each other’s audio mark…

[Tim Kolb]

Well…I think if the audio is off by less than a frame, your video is as close as it’s going to get without actually genlocking cameras so they image on the exact same clock.

You shot 30…if the clapper is already within the same frame, you’re there. While the subframe audio offset could cause some audible flanging, no human being will notice that sort of a frame clock offset between cameras after you’ve cut the scene.

Pull each audio clips toward a central sync point and edit.

TimK,
Director, Consultant
Kolb Productions,

[Alex Zvyk]

Normally that’s what I do on any 2-cam shoot, Tim, and I agree – it works just fine for lipsyncing.

But this project is different. What I have is 10 video cameras, all rolling at the same time. They are positioned along the path – arch in my case.

The idea is to replicate The Matrix effect – Bullet Time, when I’d be able to “freeze time” and move the virtual camera along the path of my 10 actual video cams.

In the Matrix, it was achieved with Canon Rebel still cams, and I did some tests with those, but it seems restrictive. For instance, you are locked to the moment the cameras captured their stills. What if the moment is off?

So I now have 10 video cameras rolling at 30fps. The idea is to sync all these videos exactly to the clap mark (audio track), and then I will be able to freely choose at which moment do I want to “freeze time”.

Problem is, in post, I can’t sync the darn clips to less than 1 video frame accuracy because of NLE limitations. And this is simply not good enough for my purpose, because in 1/30th of a second, my object (dancer) moves significantly between the cams!

I also can’t genlock those cams at the time of acquisition because those are cheap cams without timecode sync abilities. (I have to go with them due to the large number of cameras required for the effect.)

Ideas?…

[Tim Kolb]

Ugh…OK, I see.

The issue is this…those images and what “instant” in time they were “taken” is locked in.

Even if you could somehow slide them sub-frame, you can’t change when the camera took the shot…you’ll simply be moving the mismatched instances around. The images are the images at this point…

Even your initial idea about going to a 60 fps timeline would only create two instances of the same frame…with progressive, you don’t even have those two fields to subdivide that you would with interlace.

In theory, it may be possible to take the instances that you want to do the “bullet time” thing and try to Time Remap all the cameras to a very slow speed (like 25%), which will create lots of artificial interim frames, (but if the dancer is in motion, there will be unavoidable frame blur…maybe it could become your ground breaking take on bullet time?) and you might try exporting still frames if you can get the action to be relatively synchronous, and then use the still frames for the bullet time angle slide…

That’s the best I’ve got off the top of my head I’m afraid. Inexpensive cameras are liberators and hand cuffs all at the same time…

I’ll mull it some more…maybe I’ll come up with something smarter…

TimK,
Director, Consultant
Kolb Productions,

[Alex Zvyk]

The issue is this…those images and what “instant” in time they were “taken” is locked in.

Even if you could somehow slide them sub-frame, you can’t change when the camera took the shot…you’ll simply be moving the mismatched instances around.

Oh crap, you are right!

So I guess, there could be two solutions to the problem…

1. Use high-speed cameras to quantize time better (at shorter intervals), which then will allow for better sync in post. Say, 120fps camera would allow 4 times more accurate sync than a 30fps camera.

OR

2. Employ your idea to actually “blur out” the reality, so to speak, so motion flows more smoothly from one frame to another, thus actually eliminating a need to exact sync (because the difference between the frames becomes small).

Problems:

#1 – cost. I do not know of any $$ budget video cameras that record 120fps in HD.

#2 – looks like it will work, but with the obvious limitation: a very blurry motion. In most cases I don’t think 4x more blurry than normal motion would be an advantage…

[Tim Kolb]

[Alex Zvyk] “1. Use high-speed cameras to quantize time better (at shorter intervals), which then will allow for better sync in post. Say, 120fps camera would allow 4 times more accurate sync than a 30fps camera.”

This would be why this is typically done with still cameras and some sort of flash or extremely bright lighting…they can be synced and the bright light helps to capture a shorter exposure, making a less motion-blurred image.

TimK,
Director, Consultant
Kolb Productions,

[Steve Brame]

However, it looks like you’re on to something here. Innovation like this is what drives the industry/art forward.

Might want to talk to a patent attorney…

Steve Brame
creative illusions Productions

[Mark Elf]

Alex: I’m having the same issue where as one of my clappers recorded on seperate video cameras appear in the time line between frames !

I have an external mic on one video cam and on the other one I used an onboard mic. one vid cam with onboard mic was positioned a little closer to me that the other cam with external.

I didn’t expect a sound lag by a fraction of a second when I hit the clapper. the mics were maybe 3 ft apart one diagonal to the other.

As it turns out when I played the 2 videos back after I got them as cloe as possible I sure couldn’t tell the difference .

It just bugged me that there is no way to be “exact” with the audio sync and thus having the video “exact” also.

If you did find a way to stop the “snap to grid” in the time line let me know.

Thanks

Mark

[Alex Zvyk]

Mark – well… Tim nailed it… since VIDEO FRAME is the lowest time resolution unit, you are locked to what is recorded in that frame.

Sliding it would not do any good for more precise adjustments, since we’d just be moving the same visual information back and forth in time.

On the other hand, if you separate (unlink) the audio track from the video track, and change the NLE timeline units to audio units, then you’ll be able to adjust audio very precisely against the visuals.

This will not help my case (I need higher resolution time sampling for video, it seems), but it should help yours.