Fake dof with compositor???

Hey there, i’m just wondering if there’s a way to fake depth of field on video that you don’t have the source for without reconstructing the main elemens in the shot and using the z values from that to do a cheat focal blur.:confused:

DOF is differential blur according to Z depth (among other factors, of course), so in some way or another you’ll have to mask your shot elements based on depth and apply blurs accordingly. You could do some gonzo roto work and use those mattes to set up an elaborate array of nodes to isolate the depth planes, but it’s likely to look somewhat faked because there are other factors that affect the focal blur that can’t be cheated in this manner.

Fake DOF, can be done with a greyscale image, shades of determining depth, you could paint it (painful) or roto some basic shapes and keyframe, doesnt need to be tight then place them in 3D space and render that or just give the shapes a grey value material with shades of to suit some sort of depth and render that.

And the Tilt-shift effect is achieved in post the same way but the greyscale image is just a ramp across the image (or a band if you want 2 blurry bits).

“In the old days,” they would take a clear sheet of glass, smear the top half with Vaseline, and hold it in front of the camera.

You might be able to fake something with a mask … a gradient, perhaps, or just a hand spray-painted JPEG. Use this to influence a defocus or blur.

Is it accurate? “Hell, no, but neither is a sheet of glass and Vaseline.” Funny thing about the human eye, though: it sees what it expects to see, not what it does see.

Thanks for ideas and tips i’ll have to mess around and see what works best.

One parting thought: “how important to the show is this shot, or this effect within this shot?” If the answer is, “not very,” then … if you will please pardon me … “cheat like hell.” :yes:

“A difference that makes no difference” ought to be dismissed as quickly as possible so that you can focus your attention on more-important things. Star Wars Episode One shipped to theaters with a shot in which the crowd in a podracer scene was actually multi-colored Q-Tips cotton swabs. No one noticed until a “making of” documentary pointed-out the trick. (The shot was replaced in the DVD.) Since a very detailed crowd-shot had been shown a few seconds earlier, the audience didn’t “see” the Q-Tips … they saw the crowd they’d seen before, because this was what their eyes expected to see. (Plus, the crowd was not their center-of-attention … Skywalker was.)

If you’re looking for “fake DOF,” then fake is, indeed, the operative word. You want to come up with a computer-efficient way to produce a sense of depth-of-field that is simply good enough not to “abruptly ring-false” and thereby disrupt the audience’s focus of attention on what is important in the scene.

The eye is so easily fooled … because our brains are genetically programmed to “cheat” at every apparent opportunity, paying close attention only to what appears to be important and substituting conditioned images for actual perception unless something actually demands that we “do a double-take.” For example:


“Paris in the Spring,” right? Look again…


The problem with faking DOF is the same as faking any other optical camera effect – if it’s done well with an understanding of what DOF is and how it works, it’s a good fake. If you just slop some Blur on willy-nilly, even if it’s somewhat correlated to the scene, it will look faked, and break the illusion. But as always, a lot depends on the specifics of the material being processed. If it’s a long lingering shot where the camera wanders through a variety of depth planes, then it takes a lot more care getting it to look “not-faked,” than if it’s a 1.5 second rack focus that snaps past too fast to see the rough edges.

Absolutely agree. It is very much a judgment call. If you are shooting for a depth-of-field effect, a zoom or anything like that, you’re probably not going to be able to fake that at all.