Comments

Spot|DSE wrote on 11/20/2004, 5:27 PM
There isn't one. However, on the VASST site, you'll find a beautiful tutorial on how to use Deshaker written by John Meyer. It contains examples and everything you need to know. Pretty easy to do, with his tutorial.
rextilleon wrote on 11/20/2004, 7:58 PM
No plug in but Steadyhand by Dynapel works pretty good. Also Spots recommendation is pretty cool.
johnmeyer wrote on 11/20/2004, 8:50 PM
Here's the link to the tutorial Spot mentioned:

Deshaker Tutorial

It requires a free registration, if you haven't already signed up.
Steve Mann wrote on 11/20/2004, 9:13 PM
I also use the Dynapel SW. I chatted with the author about making a plug in for Vegas, and he said that if he could get the Vegas Plug-in interface specs, that the would like to make a Vegas Steadyhand plug-in.

Is the AI available? Anyone know?
Spot|DSE wrote on 11/20/2004, 9:22 PM
The SDK has been available for about 4 years now. I talked with those guys several times over the past years, they talk a good game. But don't do much about it. They need to tighten it up. Run Dynapel, then run John's Deshaker method. MUCH better results from John's workflow. Slightly slower, but good things come to he who waits.
johnmeyer wrote on 11/21/2004, 9:21 AM
Dynapel hasn't updated their consumer products in ages. I don't think they are likely to do a plug-in.
wcoxe1 wrote on 11/21/2004, 6:49 PM
I was a Beta Tester for one of the first versions of Steady Hand. I told them about the Developers stuff just before Vegas V. 4.0 came out. The corporate pres that I was dealing with said it was simply a matter of money. If they thought they would sell enough, they would make a Vegas version.

Still, no show. Very nice people, though.
RichMacDonald wrote on 1/8/2005, 10:54 PM
I think VirtualDub+Deshaker is a terrific program and is really improving my work. I find I have two kinds of deshaking I need to do. The tool works great for one but not the other.

1) You have detail in the background, the foreground is the subject but is moving around and you want to keep the movement of the foreground, just without the jerks. Deshaker studies the background and is able to detect the camera motion and correct for it. The result is your background looks smooth and the foreground is moving around, but smoother and within camera framing. A++.

2) You have little to no detail in the background, a moving foreground, and you want to stabilize the foreground. For example, tracking a bird in the sky at full zoom. Under normal settings, deshaker detects the foreground movement but considers this to be "errant" movement and discards it in its correction calculations. Thus there is no improvement. Tthere are ways to force deshaker to analyze only the foreground. (a) You can shrink the portion of the window that deshaker analyzes, thus forcing it to look only at the foreground. (b) Add a pre-filter to generate a sharp-contrast picture, where the foreground is all black and the background is all white. Or just find any filter that will remove all detail in the background while keeping detail in the foreground (*). This causes deshaker to study only the "center blob". Here is the problem, though. That blob is rarely a stable shape, so deshaker just sees a moving blob and it either performs no correction or it performs a completely incorrect correction and gives garbage. There seems to be no way to tell deshaker to keep that single blob in the center of the screem. F

(*) Its probably bad to make your foreground a single-colored "blob", rather than a detailed picture. But still, if I'm tracking a bird flapping its wings, no way can deshaker lock on its body and ignore its wings.

In summary then, for handheld shooting, virtualdub+deshaker is a great help for most cases. However, I still cannot stabilize a single object on a constant background.

Perhaps I've missed some deshaker setting. Anyone?

It actually seems to me that that latter task (stabilizing/centering a single blob on a constant background) is the easiest problem a deshaking application could have. But it requires a different algorithm than the one deshaker uses.

P.S. John Meyer, thanks for the heads-up on this.

ADDENDUM: One thing I alluded to which is a very useful technique and is not discussed in the tutorial: you don't need to run the scanning/calculation pass (pass 1) on the same avi as the final production pass (pass 2). Obviously, your pass 2 is going to be run on the clip you wish to deshake. But you can "pre-filter" this clip to generate another intermediate clip, then run pass 1 on this intermediate clip. And anything you do in this "pre-filtering" to help the calculator is good. For example, noise is bad, since its much harder to figure out "something" is moving if there is so much noise the "something" cannot be identified. Added contrast might be another; anything that smoothes out or hides the portion you want ignored and accentuates the portion you want deshaken.