All I see is the $95, 30 day dongle-based evaluation version, something about a learning version that is due out in the future, and the big enchilada for ~$5K. Did I miss something?

There is also a $1200 (retail) version that is 8-bit only.

Fusion is a great program, but takes a while to learn.

Look at Silhouette Roto to see if it covers your needs. This $280 (at Tool Farm) After Effects/FCP plug-in puts Fusion, Combustion, AE, etc. trackers to shame, and it is also available as a standalone program for more money. You can of course try the demo version before you buy it, at the Toolfarm link above.

Silhouette also has a combo Roto+Paint version for $470, also astonishingly wonderful (you may get the impression that I like this program...).

It really is a great program, I have it and love it. Sony Imageworks uses it for major Hollywood film productions too!

You can work with LUT float footage, you can vary the softness of masks for each node, the tracker follows objects like a bloodhound, even footage the others just can't follow without massive handholding, etc., etc.

Boris FX 8.0 does do corner pinning and is available for $199 to Vegas users as an upgrade from Boris FX LTD. I have not tried it personally so I really don't know how well it works.

The Learning edition is what I have. However I did pick it up on DVD at NAB, did save a lot of downloading!

So for $280 (very nice price!) I could track a picture frame in a clip, for example, and substitute another graphic (or clip?) in that location? Man, would that be cool.

For something like that you could do it in Vegas, it's tedious as it has to be done manually but if that's all you need and you want to save a buck...

Of course all these things get easier with smooth camera moves, i.e. a very good tripod.... or dollies and cranes etc!

Everyone should try this at least once in their life.

From that would come some appreciation for the poor sods in Hollywood (or nowadays more likely in Venice, Calif. ), who sometimes have to spend six weeks full time on rotoing a single scene, with only the finest tools...

:O)

It is possible to do amazing roto in Vegas by using multiple bezier masks and moving these manually for each frame. You don't have much control over the mask edges though, and pretty soon the work effort hits the level where sane people run out of the room screaming...

It is possible to save some work by not moving frame-by-frame, but instead setting keyframes on the first frame and the last frame, then the middle of the interval as needed, etc. to smaller and smaller frame counts.

Did you know that one classical remedy for camera operators to achieve smoother moves is simply a single Lifesaver candy?

Just stir and dissolve it in 4 fingers of whiskey and drink quickly.

:O)=)

Coursie!!? Love it! Love it!!! Spluttering on my early morning London cuppa coffee . . .very funny. . . .need to wipe my LCD screen down . .

Ahem - and yes, having "done" this in Vegas, it is a brain-weakening process of the highest order. If you are a control freak, then this IS the way to madness. It is not a "perfect£" solution but does work. If it is a little then OK. If you have a lot? Ugh . .. :-(

Grazie

In my "Deshaker" monograph over at VASST, at the very end, I hinted at an "off-label" application that might be let you do this for nothing. I haven't had a chance to explore the possibilities, so I cannot give you a real-life answer on whether it works or not.

Basically, the idea was to set Deshaker's controls to certain boundary limits which would force the image to stay in the same location (you can use Deshaker's boundary controls to help focus the software on the object you are interested in). You then run the first pass. In and of itself, this doesn't do anything useful. However, the vectors from the first pass are stored in a file (the LOG file). These vectors provide exact information on the motion of the object in question. These can then be applied to the mask, or whatever else you need to use as part of your rotoscoping.

Here's a link to that article. Scroll to the end and read the paragraph on "Other Uses."

Deshaker

AH! Right .. now there is something "similar" in SteadyHand. And yes I kinda remember thinking this when I read your thing John . . . Just expand a bit more on the theory and process? Please? Do I initially "set" the start around the area needing to to be tracked and then - well for me Steadyhand - will adjust through 360degs that which will "keep" the passage of that area within the tracked plane? Something like that? If this is true then I've got a whole load of "other" options I could .. if this works I think . .. well anyways .. So? John, please explain further?

TIA!!

Grazie

John, please explain further?

Most everything I post here in these forums is based on experience; from actually doing the thing. In this case, this is "dry labbed."

The paragraph from what I posted pretty much describes the process, but to reiterate: The idea is to put Deshaker into a mode where it won't allow ANY movement. You "focus" it onto some object by setting the edge cropping so it can only "look" at a very small portion of the frame. That small portion becomes the object that it is going to try to keep in one place. That object has to stay in the frame during the entire sequence. You then run the first pass of the two-pass Deshaker process. If you actually look at the video that results, it will be really wild, with huge black edges all around as Deshaker moves the entire video around in order to keep that object in exactly the same position. Not very useful, as I pointed out.

However, the LOG file created now contains all the motion vectors designed to keep that object in the same location. If the object moves to the right by five pixels and up by two, then the LOG file will have the information necessary to tell Deshaker, during the second pass, to move the video to the left by five pixels for that frame, and down by two pixels.

The key to the idea is to not use the same video during pass two. Instead, use a video that consists of your mask. It will then move around exactly in the form necessary to track the object. Of course your mask won't change shape, but you will get a very good first approximation, and for some situations, this may be sufficient.

If you know anything about AVISynth, there is a similar possibility, but using technology that does local motion estimation rather than global motion estimation (which is what Deshaker does). With this technology, every small group of pixels is tracked independently, and motion vectors are created for each small portion of the picture. This is the technology that is at the heart of MPEG-2 encoding, where you encoded one frame of video, and then create the next "n" frames of video using relatively small (in terms of size on the disk) motion vector files. This same technology can be used to synthesize new frames for slow motion (a la MotionPerfect) or to insert a synthesized frame to substitute for a bad frame.

Anyway, there is a plugin for AVISynth, written of course by a Russian (they know how to teach math in Russia), called MVTools, that provides some amazingly good motion estimation technology. It has, unfortunately, way too crude an interface to be useful by common man, but it might be adapted to provide a solution to the problem of also changing the shape of the mask as it moves. This would probably have to be done in small chunks, since any motion estimation eventually "loses its way," which is why MPEG-2 encoding includes "I" frames fairly frequently.

So, that is a little more information. The Deshaker idea is there for anyone to try and might be useful in some instances. The MVTools-AVISynth idea is way too much work for most people, and should definitely be treated more as a talking point than as anything that could be done practically.