Wow, Smoke Simulation…Across Space and Time! 💨

Dear Fellow Scholars, this is Two Minute  Papers with Dr. Károly Zsolnai-Fehér. Today is going to be all about turbulence! And we  are going to do this, and this and…this. To understand what this new paper has to offer,  we have to look at an earlier work named Wavelet   Turbulence. Apart from the fact that it in  my opinion, it is one of the best papers ever   written, it could do one thing extremely well.   In goes a coarse, low-resolution smoke or fluid   simulation, and out comes a proper simulation  with a lot more details. And, it is typically   accurate enough to fool us. And all this was  possible in 2008. Kind of boggles the mind,   so much so, that this work even won a technical  Oscar award. Please remember this accuracy   statement as it will matter a great deal.   Once again, accurate enough to fool us. Now, more than a decade later, here we are,  this new method is leaps and bounds better,   and can do 5 amazing things that  the previous methods couldn’t do. One, it can do spatiotemporal upsampling.   Upsampling both in space and in time.    What does this mean? It means that in  goes a choppy, low resolution simulation,   and out comes a smooth, detailed simulation.   Wow, now that is incredible. It is really able   to fill in the information not only  in space, but in time too. So good. Now, two, previous methods typically try to take  this coarse input simulation and add something   to it. But not this new one. This new method  creates a fundamentally new simulation from it.    Just look here. It didn’t just add a few  more details to the input simulation,   this is a completely new work.   That is quite a difference. Three, I mentioned that Wavelet Turbulence  was accurate enough to fool us. So I wonder   how accurate this new method is? Well, that’s  what we are here for, so…let’s see together!    Here comes the choppy input simulation, and here  is an upsampling technique from a bit more than   a year ago. Well, better, yes, but the output is  not smooth, I could characterize it more as less   choppy. And, let’s see the new method. Can it do  any better? My goodness. Look at that! That is a   smooth and creamy animation with tons of details.   Now, let’s pop the question - how does it compare   to the reference simulation - reality if you will? What? I cannot believe my eyes. I can’t tell the   difference at all! So this new technique is  not just accurate enough to fool the human eye,   this is accurate enough to stand up to  the real, high-resolution simulation.    And all this improvement in just  one year. What a time to be alive! But wait a second. Does this even make sense?   If we have the real reference simulation here,   why do we need the upsampling technique? Why  not just use the reference? Well, it makes   sense - the plan is that we only need to compute  the cheap coarse simulation, upsample it quickly,   and hope that it is as good as the reference  simulation, which takes a great deal longer. Well, okay, but how much longer? Now, hold on  to your papers, and let’s see the results, and….    yes! This is about 5 to 8 times faster than  creating the high-resolution simulation we   compared to, which is absolutely amazing,  especially that it was created with a modern,   blazing fast reference simulator  that runs on your graphics card. But, remember, I promised you five advantages, not  three. So, what are the remaining two? Well, four,   it can perform compression! Meaning  that the output simulation will take   up to 600 times less data on our disk, that is  insanity. So, how much worse is the simulation   stored this way? My goodness, look at that! It  looks nearly the same as the original one. Wow. And, we are still not done yet.    Not even close! Five. It also works on super  high Reynolds numbers. In other words, if we have   some of the more difficult cases  where there is tons of turbulence,

it will works really well. This typically  gives a lot of troubles to previous techniques. Now, one more important thing. Views are, of  course, not everything. However, I couldn’t   not notice that this work was only seen by  127 people. Yes. I am not kidding. 127 people.    This is why I am worried that if we don’t  talk about it here on Two Minute Papers,   almost no one will talk about it. And  these works are so good, people have   to know! Thank you very much for watching  this, and let’s spread the word together! Thanks for watching and for your generous  support, and I'll see you next time!