Welcome to our community

Be a part of something great, join today!

Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

Is there software that makes a dolly shot out of a static shot?

Stephen Pruitt

Stephen Pruitt

Well-known member
Joined
Sep 14, 2007
Messages
3,309
Reaction score
40
Points
48
Website
www.thetreemotionpicture.com
Hi all. . .

This is obviously related to my earlier question about smoothcam software, but sufficiently unrelated to warrant a thread of its own (I think).

Say we have a great static shot. Say we didn't do a dolly-in on that shot. Say we hate zooms (in camera or post). Is there some whiz-bang software that allows you to create a fake dolly shot out of said static shot by "bending" the image perspective around the objects of interest? It seems to me that while this would be very costly from a computational standpoint, it just might work. Some documentaries are doing very similar things with photographs (albeit on fixed planes). I just wonder if there is something like this out there?

Stephen
 
Stephen Pruitt said:
Hi all. . .

This is obviously related to my earlier question about smoothcam software, but sufficiently unrelated to warrant a thread of its own (I think).

Say we have a great static shot. Say we didn't do a dolly-in on that shot. Say we hate zooms (in camera or post). Is there some whiz-bang software that allows you to create a fake dolly shot out of said static shot by "bending" the image perspective around the objects of interest? It seems to me that while this would be very costly from a computational standpoint, it just might work. Some documentaries are doing very similar things with photographs (albeit on fixed planes). I just wonder if there is something like this out there?

Stephen
Click to expand...

Any software capable of 3D compositing can do this. After Effects, Nuke, Smoke/Flame - all of these (and others) have 3D image planes and a 3D camera that used correctly can do what you're asking.
 
Totally do-able. It's going to take a lot of time and experience, though.

Basically, you have to isolate every object that exists on a different depth through rotoscoping, rebuild the scene in 3D, and add in a camera movement.

Figure a 10-second shot would probably take something on the order of a couple weeks to a month for one person to do (who already knows how to do it), including roto. And that's working ~40hrs a week. That's clearly a pretty shot-dependent estimate, though.

Anything involving roto takes a long time.
 
Oh, I would add - After Effects doesn't have a real 3D environment, it only has the Nuke equivalent of the Card3D node - sort of a 3D cheat with flat planes. I don't think it would be that suitable for this kind of purpose.

And I think only Smoke Advanced on Linux has the tools to do this, not Smoke for Mac. I could be wrong though.

The main suspects for this kind of work would be Nuke, Fusion, and Flame.
 
Essentially what your asking for is the "kid stays in the picture" effect found in many current docs. This can be done with a talented retouch artist and a good compositing artist. But there is NO one, two, or even three click solution.

I've done many of these shots with Aftereffects and Photoshop. They are simple.... but time consuming.

The tut that Jake posted is dead on:
http://www.videocopilot.net/tutorial/virtual_3d_photos/
 
Actually I am going to differ with my colleagues.

You can probably skate by for shots where you wish to emulate subtle camera moves- but that is a very small set of camera moves.

Move too much and you have obvious "2.5D" effects.

If the move is that small, then I probably would ask why the heck you are moving the camera anyway.

One thing you get in a moving camera shot is a change in perspective and because of parallax you get different changes on different objects at different depths.

So, in english, what I am saying is that you can't see another side of the subject if it wasn't photographed. So, you can never dolly from an OTS to a two shot in post.

I am not telling the whole truth - you could recreate that shot, but you need a good 3D model for everything that is in the camera's view at any point in the entire virtual move, and multi-angle reference photography. Oh, you'll also need a KINGLY budget.

Even with all that, it won't look anywhere near as good as if you just moved the camera on set. It would also cost a heck of a lot more time and money, both on set and in post.
 
Charles Angus said:
Totally do-able. It's going to take a lot of time and experience, though.

Basically, you have to isolate every object that exists on a different depth through rotoscoping, rebuild the scene in 3D, and add in a camera movement.

Figure a 10-second shot would probably take something on the order of a couple weeks to a month for one person to do (who already knows how to do it), including roto. And that's working ~40hrs a week. That's clearly a pretty shot-dependent estimate, though.

Anything involving roto takes a long time.
Click to expand...

Two weeks (or a month) for 240 frames of roto is insane. Please show me examples where that is needed. The shot-dependent estimate you are working with is way off.

EDIT: See post below
 
Last edited:
Respectfully I have to disagree with you Alexander.... although your logic and facts are sound... the reality is that in 2D and even 3D effects work using cut up 3D planes or cards is a standard of practice.... watch any of the 100 BTS snippets for the VFX on District 9 and or Avatar and you will see they are placing 2D cards in 3D space.... depending on the object the parallax change caused by such moves is negligible... You also tend to have more leeway with organic objects as the human (pattern recognizer) brain can't lock on to the the parallax of say a human being and predict it as accurately as say a boxy car.

And I agree that earlier estimates are slightly off... Think in terms of hours not days. If you know the software tools in question of course.
 
A.Fetchko said:
Respectfully I have to disagree with you Alexander.... although your logic and facts are sound... the reality is that in 2D and even 3D effects work using cut up 3D planes or cards is a standard of practice.... watch any of the 100 BTS snippets for the VFX on District 9 and or Avatar and you will see they are placing 2D cards in 3D space.... depending on the object the parallax change caused by such moves is negligible... You also tend to have more leeway with organic objects as the human (pattern recognizer) brain can't lock on to the the parallax of say a human being and predict it as accurately as say a boxy car.

And I agree that earlier estimates are slightly off... Think in terms of hours not days. If you know the software tools in question of course.
Click to expand...

Thinking for example of the Video Co-Pilot tutorial linked above ... it looks like 2D converted to a move.

Also neither example represent a dramatic reframing or anything other than the most trivial example of a moving camera.

It occurs to me however that we are essentially talking about two different types of "moving camera."

Here is an absolutely artless clip I just made with my 7D. In it we go from a medium close OTS to a medium long two shot. In the process we see entirely different backgrounds, different parts of the "characters." Heck we go from having one character as a dirty OTS blur to showing them in a slight 3/4! There is an unintentional vertical move, feel free to ignore that

[video]http://www.zenera.com/movies/cantfakethis-iPhone.m4v[/video]

I don't think there is one pixel in common between the start and end frames.
You just aren't going to fake that.

So ... just move the camera already.

(I'll remove that video from my server by 4/15/2011.)

(For anyone who cares, that was shot at roughly 21mm at f/5.6 ISO 100. If it wouldn't cause me to fall off the deck I'd have chosen a much longer lens like an 85mm or a 100mm .... and a correspondingly longer dolly move, but with the same 90 degree pan.)

Here is an unrelated challenge ... make up the dialog that goes along with that shot. I've got about three dumb ideas, and maybe a fourth good one. FWIW the maquette on the right is a "female." Have fun.
 
The responses mostly focus on technical barriers: what's technically doable using current software/practices. The real question is whether this is economically doable. I can rent a cheap dolly for $50/day. Why in the world would I pay some post guru $50+/hour to create a shot that would cost me almost nothing extra to shoot in the first place?
 
I agree completely. In my case. . . the reason for my asking. . . is that I have now changed my mind on a shot already completed and wished I had done a straight in dolly movement as opposed to making it static. My question concerned whether was an economical/easy way to make this static show appear to be a very short dolly-in. It works as a static. I now think that it would work better as a dolly-in. IF it can be done.

Stephen
 
You can dolly into the shot using 3D projection and a pseudo depth map, but if its not as high resolution as say a. R3D image, then it wont work well, you'll end end up with digital artifacts. Flame has a a camera projection dolly tool that does just that, it also has advanced pixel correction to rebuild and sharpen the pixels. I don't know about Nuke, AE et al.
 
Agreed. You are limited by the resolution of the image and the type of shot you wish to create. Work within the constraints you have, or move the camera on location.

Of course it is a slightly different kettle of fish if you have three still photo's taken from different perspectives and then you can use different types of mapping like this: http://www.photosculpt.net/
 
Okay. . . this is all good to know. I'm just looking for the equivalent of a 1 to 2 foot dolly-in, so this just might work.

Thanks again!

Stephen
 
Stephen Pruitt said:
I agree completely. In my case. . . the reason for my asking. . . is that I have now changed my mind on a shot already completed and wished I had done a straight in dolly movement as opposed to making it static. My question concerned whether was an economical/easy way to make this static show appear to be a very short dolly-in. It works as a static. I now think that it would work better as a dolly-in. IF it can be done.

Stephen
Click to expand...

A 1-2 foot dolly in? That's it?

The video co-pilot link has an example of exactly that.

If the move is linear and very short, its almost certainly feasible.

Just remember that the demo/tutorial uses a static image. A moving image will take far longer. How much longer depends on the subject, your skills and the level of perfection you desire.

I frankly don't think you can alter a static into a move at a feature level of quality. (Someone brought up District 9, they used "2D cards" but they planned the moves from the onset ... its quite different.) You can get this done though for anything short of a feature.
 
Michael Panfeld said:
The responses mostly focus on technical barriers: what's technically doable using current software/practices. The real question is whether this is economically doable. I can rent a cheap dolly for $50/day. Why in the world would I pay some post guru $50+/hour to create a shot that would cost me almost nothing extra to shoot in the first place?
Click to expand...

Ignoring the fact that the OP essentially wants to revisit his decision from the shoot ...

You may have camera moves that were too dangerous/expensive to perform, so you may want to create them digitally.

You may be looking for a specific look that you can better achieve in 3D- in other words a stylistic choice. I think you may see this sort of thing in come comic book movies. That embraces the limits of the technique as a virtue though.

Still I agree in principle- for most shots just do the move on the day and get on with it. The post solution is a kludge on a good day, and often unworkable without preparation during the shoot. Its also always more expensive.
 
virtual camera dollying into a static shot is done every day. Its does not take hours like many suggest mainly minutes. To do precision dolly shots in and out and side to side, you would need an expensive motion control rig, you can just send that to a Flame op and get in done three hours for 1/3 of the cost.
 
Okay. . . anyone know a good and cheap Flame operator?

:-)

Stephen
 
You must log in or register to reply here.
Back
Top