jack frost

[MENOZ]

hello, i was reading this http://www.fxguide.com/featured/rise-of-the-effects/ in particular the part relative to the ice formation.

it says:

“After you have that wave crawling across the surface,” says Lindquist, “we then traced millions of curves through that and they were then rendered on the surface with our in-house renderer to generate the final frost look.”

so, how can I trace curves that follow the "wave" ?

and also another curiosity:

The rules for the sim were based on paintings on the surface. There would be a starting point and the painting over the frost surface would be a guide as to where it would grow.

i'm wondering if the cool leaves pattern that you can see on the ice is itself the image used as pattern, or if the paint that he's talking about is just a rough line to define the direction of the main ice formation and then the cell automation will generate the leaves shapes.

Or if the painting over the surface only define the borders of the growth maybe..

happy new year!

[pclaes]

I would approach this in a two step proces:

1) The turning on of the particles.

2) The tracing of the curves.

1) I would give each particle a resistance level based on a broader noise pattern. Then using a sopsolver I would choose a single or a few starting points. Those starting points have a value of 1, they then start to attribute transfer some of that value to their neighbours. Some neighbours have less/more resistance before turning on. You would also record which was the point number of the point that infected you. -- You will need to know this pointnumber to calculate the direction vector for the trace in step 2.

(1) Think of it for a second as if you were a point on the surface and you have a resistance of -0.4, now all of sudden there is a point next to you that is "sick", and it is starting to infect you at a rate of 0.1 per frame. If you reach 1, you are sick too and start infecting those around you. You remember the point that infected you.

(2) Now that you are sick, you are going to "walk back" to the point that infected you, and you are going to leave a little curve segment behind. You are then going to ask that point who made it sick, and walk to the next point, and the next and the next, until you reach a source point.

2) All of the point from step 1) can provide you with a direction vector "flowfield". The walking part is almost a reverse operation. You start from the end point and you work your way back to a source point.

Look for some topics here on: pathfinding, dykstra and a-star, curve growth algorithms and chaos theory if you want to know more.

[henderthing]

looks at least a bit like this process:

https://vimeo.com/25604611

[MENOZ]

hey thank you guys, your suggestions helped me a lot.

Peter my first test was only using the particle proximity to activate other particles. and just changing the search radius and the minimum number of alive particles already vides you a really nice growing structure.

I easily implemented the infection thing, that has to be the next step. "infection map" is really appropriate, i probably didn't have thought it as an "infection", this helped me to sort it out quickly.

I also thought about tracing back particles, and indeed I used point cloud to get the position of the nearest particles with a smaller "generation" attribute. It works but it generates a different structure, more like a coral, and i wanted a more uniform "dispersion".

Calculating the vector field only once on the infected points using a weighted average and using that to move particles gives a more "space filling" movement to the particles.I

Tracing particles was the heavy step, my laptop crashed a lot of times, and I managed to trace particles only for 100 frames (that generates more than 6 milions of pts.).

I still need to do more tests with a better machine, but it's seems to work very well.

I don't want to spam my reel, but.. you can see the result of what i've done at the end o this video

[MENOZ]

some guys asked me to share the hip file. I cleaned it up a bit, just deleted the not necessay nodes and here is the hip file.

I hope you find it useful, and feel free to post better solutions/approaches please!

cheers!

cellAutomataInfection.hip