[Rhett Robinson]

Shahram, without seeing the footage, my best suggestion would be to replace the video with an alternate shot, meaning – keep the audio track, but use different video, possibly cut from another part in the movie (maybe you could even do something fairly simple, like flip it horizontally, or… time reverse?). I’ve seen this done even in fairly high budget films, especially when edited for broadcast (blood & gore in horror films, usually, or brief nudity). If 1/2 of the film has a woman baring her arms, then that’s not going to work, but if you have a few segments to fix, you could cut back to a scene in another portion of the film (if it’s not repeated within a few moments, it shouldn’t be that obvious), video that matches the look and feel of the rest of the shot but really is not of consequence, like a pan shot of the surroundings, etc., or import video that doesn’t belong in the film, but that you color/grain match to fit. You probably have to work with what you’re given, so if that idea doesn’t cut it, the only other way I can think of is to simply edit out those portions, which of course, won’t work if they’re essential to the story.

Good luck… although I don’t agree with the censorship to begin with – it’s your right if you’re not infringing on someone’s copyright and distributing their work illegally.

[Rhett Robinson]

Hi Mika,
I think you mean when you scale up the layer with an applied mask, it gets blurry/fuzzy; yeah, that’s going to happen.

One solution is to click on the “continuously rasterize” button for the layer (looks like an asterisk, just to the left of the quality setting). Or, make your mask/layer larger to begin with, and work in the opposite direction (scale down as needed).

Know in advance that some effects won’t apply to a continuously rasterized layer, and will have to at least be precomposed, and there may be other heartbreakers… but it works well on masks as well as imported vector artwork. I usually scale the material outside of AE, myself, unless it’s only going to be a fly-by of the artwork.

[Rhett Robinson]

Here’s a tutorial for Photoshop ice without a plugin

https://www.alt-web-design.com/text-tutorials/ice-text.html

looks pretty adaptable to AE, with the exception of the wind plugin, but could give you a pretty good idea.

[Rhett Robinson]

Some adaptation of this might help –

https://library.creativecow.net/articles/james_colin/glass.php

I’ve used the Alien Skin plugs in Photoshop, and they do a phenomenal job, so may be able to learn from some of that.

[Rhett Robinson]

Yep – “add” works well, and you might want to play with the other blending modes. “Screen” drops out the black, but with many of these it may look a too transparent. But definitely get with it on blending modes… another way is to duplicate the layer and use a track matte – all of these are useful when needed.

As stated earlier, unmult is a free plug, and is sometime pretty darn useful… it’s used in the “Creating Motion Graphics” series by the Meyers.

[Rhett Robinson]

Not to beat a dead horse, but the time stretch (I put in 10 tracks at 10% speed) does not seem to look at every single frame as changing the FPS of the comp did, although the overall effect is the same, giving about 3 tracks per second, but doing it in 10% of the time (and reducing accuracy, of course). I then changed interpolation to auto bezier, and I think, at least for my footage, it’s a winner, at least for what I was trying to achieve. I did the tracks at 10X speed, then switched back to 100% and applied to my dots. Actually, if someone isn’t going to do it manually, it’s at least probably good enough for the witness blur for a later project.

Okay, so initial learning curve, fairly high, but the process I settled on was to just make one track after another (I was actually working on something else), then changed the time stretch back, targeted my dots (I used a blue dot for each with a number on it), and corrected the 3 that had gone astray. In retrospect, especially with lengthy footage, It would probably be better to watch it make each track, and correct if it goes astray, instead of waiting until later, as I didn’t correct until I was working at normal speed again.

[Rhett Robinson]

How is that not “real-time results”? Barring the need to replicate the children’s faces, I see no reason that video can’t be tracked at a lower FPS… when I changed it back to 29.97 it was fine. Of course, initially that took just as long, but resulted in fewer keyframes, so I did it again a different way, changing the video to be 25% of it’s original time, tracking, then exanding it back. Again, pretty good… not perfect, but pretty good, and that actually seems like a better solution than my initial idea.

I guess my comment about admitting if someone doesn’t want to do something if that’s the issue was reflective of the Tori Amos stuff from a few days ago… Admittedly, the footage is not very good. I think it’s aimed at the fans, so if you’re a fan and a video freak, maybe it’s good. If you’re a pro, and expect to work with pro material in the hopes of getting noticed in the music video market, that’s probably not the project for you. If it was, there’d be a whole lot more people doing entire projects on spec, and hoping they get chosen to be paid later. I am pretty sure that Sony is not going to post new video because a bunch of people are complaining about the quality, and I guess that was my only point – that complaining never did it for me, personally.

As far as the “commitment to dull tasks”, well, actually the stop-motion took forever, and after an initial learning period, really didn’t get any faster. I was able to dramatically speed up the masking in ImageReady, although I have, of course, learned better ways since. Admittedly, in real life, I have to do plenty of tasks I don’t want to do (like timecoding depositions), and although I’ve found ways to make them dramatically faster than I was taught, and I’ve got some ideas for more improvements, to do a “better than a computer” job, it takes a human touch, and that’s me.

I certainly don’t want to come across as a know-it-all jerk, because I’m not, and I don’t know it all, which is why I’m here on a regular basis, to soak more of it in. I see that you’re a regular contributor, and although I don’t know your background personally, I see that you try to offer solutions or suggestions where possible, and that is the reason I’m here; to learn from other’s questions, and I try to help when I can as well, but admittedly, I’m usually putting in long hours, like to actually play when I can, and occasionally, I like to spend time with a human that I’m not working with or for.

One of the many things I learned is that I’d like to be able to set up all 30 trackers at once, and let them all run (even at a heightened rate) and then correct individually as needed, but I don’t really see that happening with AE’s built-in tracker. Without testing Mocha or any other applications, I’d agree with the idea that a more dedicated tracking app would be preferable. If it’s a one-time task, AE would do fine, in my opinion.

[Rhett Robinson]

Okay, so all I had readily available to me was a crowd scene that is just over 5 minutes long… people are pretty active, but probably not running all over the place like kids will, and a couple of people did leave/enter the field of view, and the camera wasn’t locked down, like I’d want to do it.

Anyway, again, it seems it would be important to know how “perfect” the track needs to be. As was mentioned, you try to follow an earring, it’ll never happen.

In some way this thread reminds me of several things I’ve seen lately… if the real answer is that someone doesn’t want to take on the challenge (and that’s okay to say, BTW), then don’t make it based on technical reasons that don’t exist.

I agree the hat idea seems almost redundant (but that’s part of the “perfect world” scenario; the kiddos would love to wear giant fuzzy colored Dr. Seuss hats, and it would absolutely make this easier!), and I thought that it might be a situation where the kid’s faces actually have to be blotted out.

Okay, after a few trials, I set this up for an easy win, admittedly… I took standard footage and set this up at 3 (yep, only 3) FPS to cut down on processing, and made 10 tracks in a hurry, and corrected the tracks. Admittedly… this is *not* the requested scope, I don’t know what the actual filming conditions would be, or even what the use of the footage is, and this end of the spectrum is definitely just hobby for me… my money’s not made in motion tracking, and I’m more intrigued by the 3D motion tracking (like PFHoe? WOW!), but I like to be well-rounded.

Anyway, I’m glad that the original victim escaped, and I can certainly appreciate the difficulty of this task, but if it’s a one-time, one-day thing, I don’t think it would be too onerous. Then again, I’ve done off the wall things, like making a 12 minute stop-motion video, manually masked every frame of 4 minutes of video (in ImageReady… how’s that for crazy?) stippled, made 3D things in AE that shouldn’t have been, and generally just enjoy solving a challenge of something like this. Of course, I think that an initial test run would be greatly preferred to “here’s the footage, give me the product”, as I usually learn a lot in the first hour or two that can’t be taught.

I’m going to assume there’s not an award for the most verbose post… but this one has to be close. I’m glad I tested this… I did learn quite a bit in the time spent.

[Rhett Robinson]

Right, a few minutes each kid… unless this is really supposed to be exact like what we usually track motion for, you should be able to scrub through the 30 minutes & grab the track in about 2 or 3 for each, maybe less. In some ways, depending on the perfection required, you may not even want to use a “real tracker”, but follow them manually. I guess it depends on the purpose of the tracker… if it’s supposed to carefully follow head bobs to relate to autism or other subtle movements, or just the child’s location in a room. You could do that by hand in a hurry, if you wanted to. (Yeah, I know… it always takes at least twice as long as you think it will) Still, about 3 kids in, you’re going to get the hang of it, or for that matter, you could just move 30 different dots at a given interval, like 1 minute, or 15 seconds.

It really does boil down to how perfect it needs to be. If they want to determine children’s hive-like behavior or lonerism, I don’t think much is required.

[Rhett Robinson]

Okay, now I’m interested, although it looks like you escaped from having to do this. Is this related to a 1-time study of children’s movements? If they were willing to have dots on the kid’s foreheads, what about “special hats” that would be easier to track when the kids turn their heads?

Although I agree that a motion track that’s good enough to replace each child’s face would be a ridiculous task, one that is simply good enough to show their location or movement in a room would be pretty easy. Yes, you’d have to watch and adjust each track separately, but I get the feeling that this is not the kind of job that require commercial level precision, so you could burn through that pretty quickly, probably just a few minutes for each kid, depending on camera angle and where they are able to hide in the room.

Actually, I’m going to test this weekend, if I get a chance – I get the idea that AE’s built-in tracker is plenty adequate for this, just with a wider search range…