Code wiggler

[Evgenii Filippov]

Hi guys, I am trying to recreate the wiggler function (which you can find window - wiggler) to add some more features to it, but I cannot wrap my head around the math behind it. Can anyone advise in what direction should I be looking?

[Mylenium]

You would need to be much more specific and explain what you are actually struggling with. The wiggle() stuff is simply a 3D Perlin noise whose parameters you manipulate and that's it.

Mylenium

[Evgenii Filippov]

My final goal is to create a script that will have the same functions as the after effects’ builtin wiggler plugin (window - wiggler), but I want to add some more features to it, like storing wiggler’s settings to presets to quickly apply it later.

So to recreate it I am trying to understand the math behind the effect.  The plugin divides every second between 2 keyframes to N(Frequency parameter) time ranges and between each of the ranges the property’s value goes from 0 to up to Magnitude value following some type of sinusoidal path. For example between 2 position keyframes: first is at the 0 seconds and position 100, 540 and the second is at the 1st second and again at position 100, 540; I apply wiggler with freq = 10 and Mag = 20px every frame the values will go:

x: 100.00 y: 540.00

x: 109.87 y: 553.37

x: 105.15 y: 549.39

x: 101.56 y: 546.38

x: 92.75 y: 541.25

x: 98.66 y: 547.14

x: 109.84 y: 561.73

x: 112.47 y: 554.50

x: 114.51 y: 548.13

x: 118.88 y: 539.86

x: 115.24 y: 535.69

x: 108.82 y: 534.84

x: 113.02 y: 533.21

x: 111.50 y: 522.73

x: 94.42 y: 520.88

x: 90.61 y: 525.99

x: 78.58 y: 538.05

x: 86.86 y: 536.36

x: 97.61 y: 545.02

x: 103.10 y: 559.63

x: 107.31 y: 556.57

x: 108.77 y: 555.00

x: 117.19 y: 559.23

x: 124.37 y: 553.85

Yes, this looks like Perlin noise combined with sinusoid values, but I want somehow recreate the exact formula. There is also a “jagged” noise type, which I have no idea what it is

[Evgenii Filippov]

I forget to attach some screenshots to better understand the examples

[Dan Ebberts]

I think this is a pretty ambitious project, but if you're determined to try it, I think you need to gather a lot more data. I think a Perlin noise field is used to determine the intermediate keyframe values, but I think you also need to analyze what's going on with the randomization of the spatial tangents and temporal ease. If I was doing it, I'd write a little script to dump that data to see if I could figure out if it's just some random function or if it also uses a Perlin noise field for those values. I don't think you'll be able to match the algorithm exactly, but with enough careful analysis you might get close.

[thomasflensted]

There is not much match in the wiggle function? I works like this: wiggle(a, b), where a is the number of times you want your object to move per second and b is the value your object is allowed to move (both positive and negative) from the object's initial value.

[thomasflensted]

I meanr "math" not "match".

[Evgenii Filippov]

I think you misunderstood the question, I am trying to figure out the exact formula of the wiggler effect. I know how the wiggler() expression works, it has 5 parameters: wiggle(freq, amp, octaves, amp_mult, time)

[Mylenium]

You may be defeating your own intentions by not calculating the actual noise based on time and slicing up the time beforehand. That would already quantize the result and produce a different look.As per your graphs you are also not handling the keyframe interpolation and the sub-harmonies/ sub-octaves, which kind of is the point why wiggle() sometimes looks as it does when a result gets sampled. Again, I think your pre-time-slicing is the biggest issue here.

Mylenium