Cineversity

I’ve done a couple of experiments, and it seems that if I import a tracked camera (from either Syntheyes or After Effects) and place a CG object in the scene with a physical camera + motion blur, the cg object slides all over the place and does not hold the track.  However, if rendered with the standard renderer, the object does not slide and seems to sit just right.  Is that a limitation of C4D or am I maybe doing something wrong?  Is the answer to render with Standard renderer only and add motion blur in compositing?

Thanks,
Biagio

Hey Biagio,

Can you share a little scene-file? “All over the place…” sounds horrible.
Does it looks like one frame ahead, or all over the place?

All the best

Sassi

Yes, I will put up a side-by-side render as well as a scene file later this weekend.  I’d love to know if I’m doing something wrong or if this is just a limitation of the Physical camera not being able to use tracking data.

Thanks,
Biagio

Hey Biagio,

Lets make it simpler, here is a little file for our discussion.

As you might notice I have set the image-sequence (which has no “Motion Blur”) one frame “earlier”, then the cube sits on the square. (If you set it back to 0-10 you might see the “all over the place”—do I got the effect as you have it? (If you work with an image sequence, movie files seems different.)

If you have that problem, that the added object is always a frame ahead, then that might do the trick. You can switch on “Movie Camera” in the Attribute Manager—this results in shutter angle options including an Offset option.
With this offset you can then dial in the right position. All of that if it is the (more or less) one frame problem.

In short, the shutter of film cameras works a little bit different than in video cameras. Speaking of current cameras we talk here about global and rolling shutter, but the main problem is a different technology than a mechanical shutter (with very few exceptions…).

My ideal would be here a +/- 360º offset and not just a positive offset. Well, years ago I added to the discussion to have a spline/curve based adjustment, to really have all options.

Perhaps my little suggestion helps.

I have created the background sequence with teh standard renderer, to have all other parts excluded that could create a problem as well, Lens distortion (not corrected for) or rolling shutter artifacts, etc.

All the best

Sassi

P.S.: I have placed a frame number into the sequence, you might notice in the (picture Viewer) that these numbers are different to the Editor view and different to the Standard render.

https://www.amazon.com/clouddrive/share?s=dWYEOwpaRwsk4pocDiJXY8

Thank you so much, Dr. Sassi!  I looked at your file, but I’m not sure I’m getting it exactly right.  I put up this movie to demonstrate the problem:

http://www.jokeandbiagio.com/images/chromeelephantstandardphys.m4v
It has both the standard render and the physical render outputs.

Alas, I’ve headed out, so I was unable to check and see if it was just a one frame offset, but if that’s what it looks like to you, I will give that a try later this weekend.

And here is a link to the Cinema 4d file, but no worries if you don’t have time to look at it:
https://www.dropbox.com/s/2sp0lavesylaqab/chromeelephant.zip

Thanks and all the best!
Biagio

Hey Biagio,

I got your files and need some rendering- time.  I have burned in the frame numbers on the footage to see what happens.

Sequences - what I normally use - works with mov-files here a little bit different. I need some time as the 29.97 fps is normally not my favorite frame rate either.

I will check tomorrow.

All the best

Sassi

Hey Biagio,

I tried to find an easy way to explain the problem. The MAXON developer team has created a solution that comes close to their idea of shutter and motion blur. There is as well a long discussion where the shutter opens and how does relates to keyframes. The keyframes itself have a discussion on their own, are they on the start-position (time wise) of the frame or in the middle of the frame.

Put all this into an equation and the discussion about this might take a while. I do not say that anyone is wrong or right here. It was a long discussion so far, and I respect the participants a lot.

Now you might think, who cares, give me something that works for me. Right. I agree, that is the target. To gap all the problems the Shutter has an “offset” option. Only available it “Movie Camera” is enabled - hence the shutter is the core element here.
Keep in mind that you track “features” in any tracking application, and these might not always sit on the middle or start of the frame. Or do they, ...and here the discussion starts Again, it is a little bit complex. But the “Offset” should make it simpler.

In your example the movement has a lot back and fore, nothing that is easily recognizable on a frame by frame exploration: Too similar. This produces the impression of “All over the place”. If the movement would have been a left to right “pan”, with variations in speed, it would be easier to spot the problem.

So, if the offset isn’t working, perhaps the camera key-frames need to be moved (Timeline>>Function>Move/Scale) It is a simple work, after the tracks are selected. Do it on a copy.

=========

To illustrate the problem and its native C4D solution, do the following, (perhaps after you put the time code on the footage and a MG-text object connected to the frames [Xpresso])

Render a still, perhaps on frame 45, with the standard renderer. Then switch to Physical and do another render. Don’t change anything else right now. Select the Standard rendering and while the mouse-indicator is over the name in the picture viewer press the A key. Select now the Physical render image in the picture viewer and press the B key. Select the “Difference” and you might see the problem clearly. You don’t? Then go to filter (picture viewer) and switch that on, crank up the Exposure. Now you can examine if the background is the same or the object. based on the lines around the objects you can see how much they differ, and even better, same timeline slider position, but different frame numbers (!?) yes, that is the way to solve a problem but to create a new one,  certainly a difficult balance the developer had to go here.
My personal take here is to try to keep the frame numbers in sync, if you go further in VFX work. If that is final and only grading will happen to the the material, then it is not important (not really)

A little hint, the less the back ground shows up, the more in sync it is. That might be clear - but just saying.

=========

So, it is a little bit difficult to dial in the right offset (perhaps in combination with shifting keyframes +/- one frame). But if you establish a working Pipeline, this is the work that needs to be done one time, then you know the “parameter”. A reason why I normally like to stick with my apps and not run and gun with any new cool toy. As a side note, not based on your lens, it seems nearly distortion free in the shoot, I wouldn’t like to work without the lens workflow during tracking.

I hope the images below give an idea of what I’m taking about. Yes, it is a little bit abstract. I used only the shadow disc and you might see the difference in position between the Standard and the Physical render and the different frame numbers. In the other image I changed some of the details I mentioned above to get it more in sync with the standard. It is a puzzle.

All the best

Sassi

Wow!  Thank you so much for taking the time and going into such great detail!  I THINK I understand it, and will open my scene when I get back to the office Monday and follow the steps exactly as you laid them out.  I can see from your reply that there is a logical explanation for the behavior, and it seems that means there is a solution as well.  I would very much like the option to use the physical camera when match moving…though I have to admit as I sit here binge-watching your JET tutorials, it seems that great results can be achieved without GI, and by mixing scene motion blur with RSMB.  By the way…is that still your preferred method for getting motion blur, or do you find yourself using the physical camera for motion blur these days?

Thanks again so much!
Biagio

Hey Biagio,

As usual, a problem comes as often back—until “you” solve it.

There is some beauty in the RSMB (but also some limitations) and I like the guys a lot who did it, they even got already an Oscar for their work.

The Physical Render is nice, but (of course) much slower. It prevents to a great deal Multi-pass work, as the “data” products (Depth, Motion or Normal pass, etc.) wont work anymore (either way). On the other hand, it is a large problem to get motion blur and depth of field correctly in 2D compositing—at the same time together. The Physical Render is great in this discipline.

I have just stopped* (for tonight) tracking the scene in Syntheyes. (*I saw the forum had three new posts). I like Syntheses and know the app better than any other tracker I have learned in the past ten years. So I need to do it one time at least to see a how it works in this case. I would love to find something that I can tell, like a formula. My last test was to render out (Ae) a sequence, to get this 29.97 fps stuff excluded from the equation. No idea if that might cause a problem, but that is the way I check things out. Exclude everything than use parts again.

All the best

Sassi

Hey Biagio,

I have set up the shot as an image sequence. So I have no problems to set up the project as 30fps vs 29.97fps. I know it is only one frame every 1000 frames that could differ. I work in 24fps and feel more comfortable in this, while exploring. Will say, exclude any potential trouble sources to see the the core of the problem.

So I have carefully tracked the clip in Syntheyes (just to haven an option for the lens distortion, which I miss in Ae’s native tracker. Even if your shot is nearly free of distortion, again - exclude all problems to see the core problem. I had an Error of 0.502 hpix, which I think is acceptable given HD and some noise in the footage. The footage had very little parallax, and initial top to start position move would have helped. ;o)

Sorry for the long prelude, but information is everything for tracking.

My best results were given with the standard render with no movement of the object on that “chair”. This was my reference to anything in the physical render.
Just switching to physical showed everything that I had initially ask. So a comparable solution/set up from your problem.

(This reminds me a little bit on my preparation work for a pipeline or a single tutorial preparation, I like to know it in depth, I don’t trust 3rd parties)

I did all the default rendering with the frustrating results, just to proof that point again.

Then I moved all camera keyframes (I used a copy of course) one frame ahead. Just to be clear - the initial frame has no animation then. (But you are a pro and I assume you work with “Head” and “Tail” always, so no harm done.

I switched the camera to movie camera and 180º shutter (the standard for film around nearly 100 years, and the offset to 180º with an efficiency of 70% (I argue this value, as it seem not to be correct, but hence it is the default, and I left it there for the time being.)

I can check the footage at 200% and can’t see any slide, well as the hpix error mentioned above, but I haven’t undistorted the footage, and I got this result close to the frame. I blame this tiny bit on lens distortion, which is very very small in your case. :o)  Perhaps your shutter was different? (I have the feeling I’m pretty close to that) All of that can change the “integration” work as you know. It needs to match.

I will render some tests with other offsets. If something dramatically comes up with this test, I let you know.

So, time to start my Sunday :o)

Enjoy

Sassi

P.S.: If you do more camera tests, perhaps you move and pan the camera, then stop and pan further, in that way you can see more precisely the amount of offset you need.

Here is the file, as discussed in the last post.
Camera keys one frame to the right, and Offset 180º
I rendered the frames 0-111, I think that should show it.
Especially the last frames, where it is really fast moving.

https://www.amazon.com/clouddrive/share?s=uG87Eb2US0UgQNK02Ig5Z0

:o)

Whoah! Just saw this. Thank you so much, Dr. Sassi!  Downloaded the sample movie, followed your directions, and voila!  I now have a decent track!  All of your tips are very much appreciated, and I’m truly excited to now feel that i have the option to use the phsyical camera in a tracking situation when the situation calls for it. 

Really appreciate you taking the time…sorry for the delayed “thank you!” but I just noticed this.  Have a great day and once again thanks for being so helpful to all of us.

Best,
Biagio

You’re welcome, Biagio :o)

I hope the offset value will be increased in the future, so we have it a little bit simpler ;o)

The feeling of a stable tracking is very rewarding, isn’t it? It is certainly a key feature to have.

No worries about any delay, I know how busy you are.

All the best

Sassi