DIY timeslice effect - MATRIX
 

Ok, I want to build that rig, I have budget of 50K-70K to do so. But am not sure which way to go. If any of you have suggestions, I would appreciate your contributions very much.

I know basic principle about it. I know that I have to syncronize cameras to shoot simulatneously and so on, but have lots of uncertainties like.

Which camera to use for example, is the biggest question. (I want to go DIGITAL way)

So please share any ideas you might have.

Just in theory for $50k one could buy something like 60 DSLRs with kit lenses.
Maybe Canon`s Rebel XT (350D) and shoot those in raw, 5 frames in ca 1,5 sec.

Raw files from those cameras have quite good dynamic range for that price.
White balance is also fully adjustable in post plus aperture is adjustable +/- 3 full stops.
RAW workflow in RawShooter for example is quite easy, cheap and fast.
One can even pan and zoom those shots as 8MP images are quite bigger than HD.

Imagine maybe only 180 degrees of movement with 60 adjustable cameras on ball heads assisting two main cameras directed in opposite directions of subject movement and all that on a moving platform - oh my what kind on freezed twists and spirals one can do :)... Output from such a rig could pay for itself quite quick I assume.

Some additional ultra slo-mo morph in Shake 4 and you`re done by my quess.

Freeze
 

Check this out:

http://www.whites.com/heads_cranes/big_freeze.asp

there are 2 things, that I know of, that cameras should have:

1) external triger
2) manual focus and iris

what else is important? What cameras will you recommend? with the lowest price of course.

have anyone experimented with it?

just go with Canon Digital Rebels - if you buy 60 you'll probably get a volume discount.
you'll need to connect them all to trigger them and they use a 2.5mm stereo jack type connector which you'll need to figure out how to make an adapter that can go to all the cameras.
After that, I think buying a railing or something you can mount them all to along with a cheap ball head for each camera should work well.
Stick with the kit lens which ships with them and you should have a great rig.
As previously mentioned, you can rent the whole rig if you want to save money.

Thanks Andrew,

So you recommend rebel xt? I will check it.


Ok, also there is another problem I thought of, the sleeping mode that all of the cameras go to if you leave them untouched for some time. If I can not turn the sleep mode off, then I will end up, having the first camera being off, untill I even turn on rest of the 60 cameras on. Unless I hire 60 people :*)

So what do you know about that?

Thanks for contibutions. After the rig is ready I will share the experience.

also, what do you think about nikon d50?

it is cheaper than canon rebel. But maybe it has some disadvantage? do you know? does it have manual aperture and focus? and external triger?

The Nikon D50 is another excellent camera. I'm just not as familiar with it as I am witht eh Rebel XT. The D50 along with any Digital SLR camera will have manual aperture and focus.

Now, on the Rebel, you can disable it from shutting off after a certain amount of time so it will always be on. I don't know about the D50 and I don't know if the D50 has the connection for remote triggering, so you may want to look into those things before deciding. It should be on Nikon's site - I'm kind of in a rush now, otherwise I'd check the site and look for that info, but it shouldn't be hard to find.

I created a very low-budget version of this effect for a client a few years back. Not nearly as polished as the kind of thing you see from the guys at timeslice.com, but it was effective for this application.

The production was a martial arts training DVD teaching wood breaking techniques and we used the effect to "freeze" the instant when the wood snapped into the air. We used video cameras so the "sync" was done later in post. With your budget, you could probably purchase a number of very good DV cameras for this use. Color correction and stabilization is MUCH easier these days since there are a number of software tools available specifically for these tasks.

Anyway, here's a write up on how we managed to pull this off a few years ago on an almost non-existant budget:
http://www.jushhome.com/nick/Tutorial/freezetime.asp
(this article has now also been published in the O'Reilly Press book "Digital Video Hacks")

The article only links to a 21Meg WMV example, so here's a 9 meg QT example file as an alternate:
http://www.jushhome.com/nick/media/nbbfdemo.mov

Hope this helps.
Good luck!

The other important thing to remember is that when this effect was done for matrix they needed to use pretty complex morphing technology to smooth out the movement in post. Even if you had 60 cameras at 30 fps you would be limited to a 2 second move, trust me it is going to require some post work. The other option you have is depending on your subject matter is doing it all within 3d. Obviously if you need this effect done repeatedly or in dynamic settings (ie someone splashing through water) this would be cost prohibitive. If you only need the effect done once or twice with only a character or two being the only 'dynamic' things in the scene, duplicating them in 3d and doing the camera move might be the way to go. Like any type of special effect, the specifics of your rig, and ultimatly the solution to getting the shot that you need solely depend on the specifics of the shots that you have to get.

BTW: Here's a great article and podcast interview about how Mark Ruff achieves this effect with near realtime results using arrays of Canon 10D cameras, each linked to their own macmini:
http://www.fxguide.com/article323.html

Mark's site is http://www.ruffy.com

Great link, thanks you Nick.

I was going to say do it in 3d, but Rob beat me to it. for 50K you can get some serious 3d artists.

Anhar

So, here is an update, for anybody who is interested, I finally got my budget and now I am in process of order. I decided to go with canon 350D as you recomended. I hope everything will be fine. Now I am trying to solve the problem with how to mount the cameras to the rig. I mean, I need some heads, designed for this camera or can I make them myself? any suggestions? what will be the best way to go?

And, thanks for the help again.

Just found this thread. Levan. To get smooth movement from one to another, without morphing, you could try slowing the shutter, and overlapping integration, so it would appear to be smoothly going from one frame to another.

Have fun, look forward to seeing the result.


Re-edit:

Do it in 3D, that gives me an idea. If you can scan the objects into a 3D model, then you can do it (but scanning and motion capture might be costly) and you can also motion capture movement of each object or all together (and video) and apply it to the captured model with artists filling in the details.

But an interesting one is that 2D pictures contain 3D detail, I understand that Intel now has software to make 3D models from 2D pictures, which might help filling in the rotation in between frames, even reducing the amount of cameras (but reducing cameras, eventually it has to reduce quality).


Wayne.

Hi Wayne. Thanks for the hint. What about that software you mention? I will not reduce the number of cameras, but the software could still be usefull. :*)

That's the thing, I can't remember, but it is bound to be on some Intel site, try their search functions.

Re-edit, just looked for an hour or two, haven't found it yet, but was Intel based.

So, I have ordered the cameras, and it will be with me within next 2 weeks. Some guys are building a sync system. For the begining I will only be able to triger all the cameras at the same time, but they will develop the software so that I could trigger the cameras with selective time delay.

Now I have to think about the rig. Do you have any suggestions? How to build it? I was thinking about buying some support heads and then build a rig where I could screw this heads. So, if you have any ideas please let me know :*)

The guys from sequential pictures did a send up of the matrix with some slo-mo-ish low budget work in it...I think they just invested in a tennis ball and some string...did all the background replacements in Photo shop, but that could as easily be done in after effects.

on the mac, Morphx does a good job for free, and I'm sure there's a PC morphing programming that will do a good job for free as well...I took the frames from the board breaking video up top here...looks really good about 30 mins of work to morph the first 2 frames convincingly in free software.

I recieved all 60 of the cameras. I did the syncronization and it is pretty nice in this aspect, but I have problems with aligning cameras in the right way. The rig. What do you suggest? even slight and unnoticable difference between cameras, makes the final sequence shake a lot. I will upload test clips and look at it and please let me know, how I can solve the problem?

I will appreciate your suggestions, and as promised at the end will share the whole experience so that anyone in the world can go and do it.


Download video (only 1,6 mb quicktime) http://rapidshare.de/files/20500610/3D.mov.html


EDIT: If the question is not clear, the problem is about aligning the cameras in the correct manner. I have tried several things but it just doesn't work. Well I am improving, but I really need to understand in which axis do I need to allign most of all.

Hey, good job, looks nice to me.
But heres an idea maybe...could you use like a focal point like lets say painted line on the wall, then use some yarn or string to "map" out the angle and such, sorta use a guide rail to align the cameras. I'm sure this could be solved with math, but I'm an art major, so I don'tknow math.
Maybe you need mounts that are measured of in degrees, then turn each mouth X number of degrees as you go down the rig?

But seriously, it looks good and i think it passes fast enough so that any miss alignment is not to noticable.

(tagged on re edit)

I actually know a bit about 3-d. So let me clear up a few things about 3-d software. Motion capture is usually something you do that involves a rig and special software and maybe even ping pong balls. Rotoscope is where you use pre-exsisting images to figre out motion. (like what bakshi did in wizard or lord of the rings or what their doing in a scanner darkly)

You can scan 2-d images into a 3-d file, but the results are not somehting you'd really want to have to work with. Usually you'd have some animator sit down and re model the scene from refrence photos and useing known distances.
This also depends on the effect you want. Doing this in 3-d, liek say with a 3-d package like maya would be a fairly easy taks, but I actually like the way it looks now, makeing a leap to a 3-d enviroment wold be sorta a shame.
I also don't think it needs to be morphed really, unless that's somehting the client demands. It's cool the way it is!

how are you viewing the image you are lining up? If you have an lcd, you can use saranwrap and a marker to mark a point you align on each cam.

afterwards, use software to do a stabilization.

I do not have time to read the thread again but I think I have the gist of it.

What about a arced flat (ribbon) rail with slot through it on one side, 60 precise mounts to mount the cameras flush on in exactly the same angle each one. The mounts have a mount that slides in the slot with no play, they slide up against each other so the mounts are flush against one another. Because the rail holds the desired curves and the mounts are all the same, they naturally space the cameras and cause the angle of the cameras to follow the curve of the rail and touch flat against each other. The rail could be made of an material that can be tensed to produce a curve or shape, but realistically such curves will tend to be parabolic rather then circular.

The other is to buy something that has holes/mounting points and use those. Some sort of strip/rail might have that (hardware store supplier might be good place to find experimental pieces).

for the cameras that slide through the slot and

for stabilization software, shake does a good job ($$$) or cinelerra (free). You could do it manually positioning the frames individually, or slow it down a bit more up front by tweening with a morphing software (plenty of free packages)...I'll try to whip up a quick example from the file you provided...Version tracker has MorphX available for download (free) to play with.

A lot of your "jitter" is coming from color balance inconsistancies from camera to camera. Equalize these and you will be cleaning out a substancial amount of the multicamera artifacts you're experiencing.

To help remove X/Y jitter, us a stationary central target (ping pong ball on a stand), shoot a sequence and then run a 2D stabilization on this in post (After Effects Pro, Nuke, Shake). Then, you can copy the stabilization data to your followup real shots. Note, that the most obvious positioning jitter come from sudden shifts in camera tilt (clockwise, counter-clockwise).

Yes, you are correct. Even thou, I only read this post today, yesterday after a lot of experimenting I came to the same conclusion. It appeared that the marked center in the camera viefinder doesn't match exactly to the center of the taken picture. So what I did, while aliging cameras to one marked point I then shot the point from all the cameras, and then I removed the mark and did my shots. Then in AE, I have aligned picture to the marked point in the picture and copied the corrected keys to the actuall shots, and the result are great. I never tried to correct the colors, and that is not going to be a problem, for this moment, I just wanted to achieve the stable XYZ movement and I did it, even thou it took almos a week of experimenting. Today I will be doing a first fully corrected shot, and tomorrow will upload so that you can see and comment.

Thanks for the help anyways.

Hi, again. Please help me here again if you have any idea.

So, I did alignemnt thing and the motion I get at the end is smooth and very nice. But what happened to be a problem now is sync. there is a slight difference between times the cameras take pictures. I sync them with 2.5mm jacks set up in parrallel connection. After I didn't get the good sync, I did a test. I aligned 15 cameras to the stopwatch and did the shooting. my stopwatch was as detailed as 1/1000 of a sec. So in some cases all the cameras take the same moment but in some cases there is a difference in 1/100 of a second. I couldn't see the pattern and logic, but it is 50/50 sometimes it is sync and sometimes it is not. I have no idea why. Maybe any of you have an idea? All the parameters are identical in the cameras and are manual. also, there is no pattern in the cameras. I mean, I couldn't single out a camera which has delays, it varies from shot to shot. also, I found that in most cases there are two positions in time, some cameras take one and others take second, so there are no multiple differencies.

So any ideas?

I was hoping somebody else would help you, because I am getting away from the lists for the moment, and I don't really know the answer.

I forget your setup (I haven't read through it for a while) it could be in the timer that generates the signal, it could be in the propagation of the signal, it could even be length of wire (electricity travels ten or 100 times slower then the speed of light normally) but yours changes, so it won;t be that, unless their is a resistance/capacitance problem with wire length that effects the way the trigger circuit responds to it. It could be the trigger to the signal not functioning proper, or normal noise problems in the trigger circuit. It could be that the signal going to all the cameras isn't truly parallel so gets out of sync or that the cameras themselves have some variability in their operation that causes the timing to come in late (or early) likely. But there is always a factor of error (unless a timing circuit is 100% accurate, and I don't think we can do that) but it should be accurate enough to constantly be within a hundredth, thousandth, or millionth of a second.


Glance at it with an expert eye, and test the most likely and the simplest to test, potential problems first. The electrical problems would require electrical equipment to test but it should not take more than a couple of hours to go through them all) and if there is not any obvious causes then it is unlikely to be them, unless you have eliminated the other possibilities (electrical problems can propagate, if the camera timing signal runs sequentially from camera to camera. Contact your camera company to ask expert technical people about internal camera timing problems explaining you have bought all those cameras and it isn't working. I think unless there is something not setup right, it might likely be internal camera timing. If they have a service/technical centre nearby maybe they might like to come and have a look.

Sorry this is not more concise, but it's after 3 am here.

Anyway, I am trying to get away from the forums. Have a look at my DIY projects Digital Cinema camera summary, you might find it interesting for when you finish your current jobs ;).

http://www.dvinfo.net/conf/showpost....&postcount=470
http://www.dvinfo.net/conf/showpost....&postcount=471


Thanks

Wayne.

Wayne, I really appreciate your attention and contribution. I realise that the problem I stated was very hard to solve on a distance. Even thou you there is no real way to know, you have tried to help and I appreciate that. I think I found the cure thou. First, I contacted www.ruffy.com as they also use digital canon cameras and thought they could help. Someone replyed saying that canon d350 is not good enough for the sync and there is no way I could solve the problem. That sounded so unhelpfull. After testing everything that came to my mind I have thought about the mirror. Because shutter is a very accurate mechanism, the error in the shutter was almost excluded from my list of potential problems. But then I remember the mirror. Because mirror has to go up beffor the shutter triggers. And there is no need for accuracy in the mirror, the only thing the camera needs is to know when mirror is up to trigger the shutter, not more than that. So I went into the menu of all the cameras and set the mode where on the first push of the button the mirror goes up and then on the second push of the button the shutter is triggered. ( that is used in marco shootings, so that camera doesn't move during the shot ). All cameras I have aimed to a timer with 1/1000 of a second. Before I set this option every 3rd shot I made had some difference in at least 3-5 cameras, after I turned the option, I have done 20 shots and 19 of them were accurate and syncronized to 1/1000 of a second. I don't know what caused inaccuracy in that one case from the 20, but I thinks that I have solved the problem. I will be doing some test tomorrow, and see the acctual result. Promise to let you know how it goes.

Hope this thread will help anyone in future who will try to do the same as me.

Thanks Wayne.

:*)

Sounds like what I thought, you could discuss this with the camera company for a firmware fix and get 20 out of 20 shots in sync. The problem with my last reply was that I was so tired, that it is hard to narrow the analysis.

Did you see the ideas I posted for using multiple still cameras to produce HD video? People sometimes ask about making still digital cameras do HD video, so I thought I would post some fun ideas that they could play with.


Thanks

Wayne.

I just discovered these links that might be of interest to you.

http://graphics.stanford.edu/papers/highspeedarray/
http://www.dvinfo.net/conf/showthread.php?t=62933

www.framefree.com
 

I recently stumbled across this morphing software and it seems to me it might help you in your project.

www.framefree.com