Leopoldo Perizzolo
Forum Replies Created
-
Yes, you’re right about functions. I forgot that we weren’t talking also about memory consumption, since it’s irrelevant in such applications. Yet, functions can still be useful when you do have repeating sections of code or to make operations clearer to understand, and also because when you have to apply changes to a piece of code, since you don’t have duplicates, you only have to do it once and you get a lower risk of errors.
Leopy
-
If you have Element, you might do something that Andrew Kramer did at 6m 53s:
https://youtu.be/vIEdhlS-zYA?t=6m53sLeopy.
Some contents or functionalities here are not available due to your cookie preferences!This happens because the functionality/content marked as “Google Youtube” uses cookies that you choosed to keep disabled. In order to view this content or use this functionality, please enable cookies: click here to open your cookie preferences.
-
In cases like this one I would recommend a 3D track of the body (the torso should be sufficient), basically using some advanced software such as PFTrack or Boujou; even if you’re on a low budget, you may be able to download a free trial version that lets you track for a couple of weeks. Otherwise, you could try AE’s camera tracker in reverse (see this link from VC at 6m 53s: https://youtu.be/vIEdhlS-zYA?t=6m53s, although it usually refuses to solve the camera in very low contrast scenes, without markers or if the subject is far away…even more if not even Mocha can get a good track) or as a last resource a 3D manual track.
Since all those methods work in 3D, you can use some sort of fluid simulation and get a realistic result even when the subject is turned sideways from the camera. If you don’t know how to work with simulations, you might get away with just multiple layers stacked at different angles.
If you absolutely can’t pull off a good track, a simple solution might be to fake it. One way would be using a lot of motion blur, another speeding up the scene (as long as it preserves a natural movement) with a bit of forced motion blur, or maybe by pulling a rack-focus as soon as the subject gets hit.
The very last suggestion is to re-shoot the scene. Often times it’s the only way to solve this kind of problems.
There’s no magic in softwares, if you can’t tell by yourself what the movement should look like, there’s no way a program can do a better job than yours. Sorry, that’s just the way it is ;P
Hope you find out a solution.
Leopy.Some contents or functionalities here are not available due to your cookie preferences!This happens because the functionality/content marked as “Google Youtube” uses cookies that you choosed to keep disabled. In order to view this content or use this functionality, please enable cookies: click here to open your cookie preferences.
-
Take a look at this video:
It should pretty much answer you question.
Leopy
-
Maybe I’m a bit late….anyway:
1st effect:
– go to Layer -> Autotracer -> Enable preview -> mess around with the values while looking at the comp and seek for the creation of a mask all around the hand, keeping the “minimum area” almost as high as possible (to avoid double-masks)
– apply VC saber on the newly created layer with the mask (it’s a free plugin) -> select the source as the mask path on the effect -> refine other parameters as needed; you should get a nice glow all around the hand
– since you have the mask layer, duplicate that and a copy of the hand footage, then set the hand layer as alpha (or alpha-inverted) track matte of the mask layer; this way you can control the hand separately from the background and apply an exposure and glow effect.2nd effect:
– do basically the same thing but this time apply a more subtle exposure and glow to the layer with the person.Leopy
-
You have to use expressions to adjust as needed; I would recommend setting it as follows:
1. Set both values of the 2D position point to 0px;
2. Paste the code:
// assuming you want to paste this on a 2D position point
[value+thisComp.width/2, value+thisComp.height/2]You can still edit both values, and every pixel you set is a “displacement” pixel from the center of the composition. Try.
Leopy
-
There’s a pretty easy fix:
take the layer that obscures the thing that you want to see and set it as an alpha-inverted track matte; basically, when the layer you want to see is covered by the alpha channel of the matte, it will be obscured. You get for free the fact that the matte goes automatically transparent ????Leopy
-
As far as I know, there are no easy ways to paste position data to a mask path, also because its position depends on the layer on which the mask is applied.
What I suggest is to use a track-matte (“subject” being the girl, in this case):
– create a new completely white solid;
– set the opacity to 20% so that you can see what you’re doing;
– create a mask of the subject you want to see, being sure the mask is applied onto the white solid and NOT on the footage layer;
– parent solid to null, then scrub through the timeline to see if the solid moves so that it covers the subject from beginning to end of the composition;
– set the opacity back to 100%;
– assuming the white layer is on top of the subject layer, set the track matte of the subject layer to either alpha or luminance with respect to the white solid (press F4 if you don’t see the switch on the timeline).That should do it.
Leopy
-
AE computes only the value of expression that have not been calculated yet; after all expression of a specific keyframe has been calculated, it stops computing and outputs the value (unless its value depends upon other frames of the comp, such as if you use valueAtTime() ).
When an expression is turned off (the little Equals ” = ” sign becomes Not equals ” ≠“), it’s no longer being processed by AE; you can get a proof of that when you accidentally have an error such as a division by 0 and the expression gets turned off automatically: the project file does not freeze trying to compute what the result might be, as it’s being disabled.
I don’t think there are lots of trick involved…the basic rules of thumb are:
1. keep you expressions as compact as possible (meaning try to avoid repeating sections of code that change very little), which leads me to point 2:
2. use functions whenever an operation has to be shared across multiple iterations of a piece of code;
3. some methods are more CPU/GPU intensive than others; one of the most intensive being sampleImage; be sure not to have multiple layers or expressions that use it.
4. generally speaking, “if” statements take less time than math functions; that is, unless you have to call bigger functions to compare two or more values while you could use already calculated variables.
5. speaking of multiple if statements (e.g. if((A && B) || (C && D)){} ), try to use the java “short-circuit” capability: basically, in every if/for/while statement always put first the comparison that it’s more likely to prevent the if statement to be executed (e.g. if you know that the value G is 99% of the time = 0, a if statement should be constructed as follows: if(G == 1 && banana == 1); since G is probably 0, the program will skip the second comparison).I think that’s it! Hope it helps solve your problems.
Leopy
-
The box effect is probably caused by the layer on which you applied the effect being too short in height; try applying the effect on a bigger solid layer, so that the boundaries are out of the comp and not visible.
Leopy