I was confused. This is U+26D3, which according to Unicode is "chains", i.e. "tyre chains required", but it is rendered as whitespace in my browser. So, this post appeared as a very tall post of absolutely nothing…
@@LevantWasTaken i dont think it has to do with fun but rather the way information gets delivered.. hearing information will always feel longer than if u actually do something but hey i might be wrong maybe he didnt pay attention and was looking at the clock then ofc its gonna feel long
@@1ucasvb Everything in this was incredible. I kept thinking the towel simulation would fail eventually due to the huge forces, but it just kept working...
the towel is sooo realistic btw we humans with our big brain can simulate things like towel twisted. it might look crazy but we do it using data learning i wonder if those algorithms do it with pure physics calcul or if they use memory (which is kinda overpowered as we can barely align multiplication while we can predict the behavior of flowing water)
Wow the twisting blew me away. I have no idea of the physics behind it but even just the fact that it retains its volume as it forms those extra knots is mindblowing
Hello, I am just a French spectator with very little knowledges of computer graphics but I just wanted to congratulate you on your work and your popularization work. I find putting subtitles on every video extremely cool. Being very interested in SIGGRAPH, being able to have access to so many technical details on papers so little known to the general public is simply great! thank you very much !
I am glad you are referencing houdini because it is important to show the comparison for what tools we already have implemented for artists and how much further the research has come. I always find myself with a sense of envy watching this channel because the simulation science evolves so fast but it actually gets implemented in the software so so slow. You keep saying 2 more papers down the line, and we see how far it comes in 1-2 years, but even after these things are demonstrated at siggraph it will be another 5-10 years before they become widespread with actual use. It makes me sad because we have such great researchers but not enough software developers who are familiar with these papers to work with them.
This is mad bonkers the amount of detail that comes in the material displacement without mesh tangling! Also a few minutes per frame was much more reasonable than my original thoughts. Still a long cpu time, but hey, the results speak for themselves
Man, the first time I awe at the realism of these simulations. If they add on textures and do not tell me that they are cgs, I don't think I'm able to tell them apart anymore. Absolutely brilliant!
I really liked that different format of video because we can see multiple advances in the technology industry in one single video, even though they aren't so detailed it's a cool thing :) I think it would be cool to the channel if it became like a 15 day-ish basis series, or even monthly, maybe they could be like some bigger-ish videos compared to the other more usual u do, I think it wouldn't even be a problem if it includes contents that you already talked about in specific videos 'cause some people don't have so much time to watch and it'd be a easier way to it keep up date! maybe it could turn into a "top 5 papers of march", Idk, think it could be great! although they probably need more time to be made I think it'd be worth, hope that idea get to u, love your content
So the research advances each month by what I'd predict for the next couple of years. This is astounding and I can't wait to see what advancements will be made next. It feels like soon there will be no problems left, other than making all of this real-time and building bigger and bigger scenes (although I know that isn't true, there is so much interesing work to be done, but you know, every video feels like this is it, we've reached the future). Greetings from Göttingen
Software engineer here. 776.2 seconds / 264k average 'contact events' == an average of 2.94ms per 'event.' This is insanely slow. and remember, this is on an 8 core 3.0ghz cpu. It should be churning through these several magnitudes faster than it is. Seriously, anything that takes even a faction of that much cpu time *per calculation* needs some serious attention. (This is sadly pretty common for academic code as in that field performance isn't even a criteria, let alone the primary one.) He said it would need to be re-implemented to run on a gpu, so this likely means it isn't tuned for parallel processing. Honestly, it likely isn't using multithreading at all, so we can assume it's running on a single core. (aside: good choice of Intel for this, as single-core is the only thing they do right). Best case, though, that's roughly 8.8 million cpu cycles for each contact calculation! (which only includes 1/264000th of the cycles for the full render code). And even if I'm wrong about the multithreading... that number only gets worse the better the code is at it. I am totally impressed by the dev(s) getting the simulation so perfect, though. It is a joy to watch. But to call its speed impressive? Just no. It needs some love and attention, including a serious refactor.
@@sagemeline Thanks for the info! These are still some impressive simulations, but it's good to know that they can get even better with more effort put into them.
I wish you could go into a little bit more detail about the differences between the new simulation methods and papers and their previous competitors. Great video!
I remember doing that to rubber bands as a kid, if you get the knots to go along the whole length, it then does a new top of knot made out of the curled up knots
4:46 What is even more fantastic, is you Károly. Your videos are really charming. I'm always looking forward to more science and technology breakthroughs with you.
idont know im just a beginner, my opinion is that no matter how many times simulations can make the results to a real world the problem is only in air particles and gravity because the earth has a round shape then it also affects the real physics. it's the same as you do it in an space but in all my opinion your guys work so impresive.
@@TwoMinutePapers There is an old video out there (ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-AkB81u5IM3I.html). I think it shows exactly the same clip (look at the background, its the same). In the old video it is stated that a golf-ball was used. This may be a big conspiracy :D
me watching the video: "of course, there's a lot you can do with a gpu" Dr. Károly Zsolnai-Fehér: "...all this is performed on your processor..." *Dolphin in Town by Kingo Hamada starts playing
and later: "... minutes per frame ..." This is horribly unoptimized. It uses (776.2sec / avg 264k calcs) or an average of 2.94ms of cpu time per calculation on the cloth folding simulation! Holy crap that is slow.
My computer cant even handle the motion of a liquid going through a pipe and you guys are showing such crazy stuff. I bet they did this on supercomputers.
Wow! That’s insane that there are never any self intersections especially in the twisting simulations I’m a 3D design engineer by profession so I can appreciate just how incredible that is!
And now imagine this kind of accuracy in physics accelerated as approximated AI/NN models that can easily run in real-time games on most modern gaming GPUs. I can't wait for the first games with proper physics like this, especially in VR; what a time to be alive
So now we've unlocked the ability of "the" Simulation to create accurate simulations within itself. Next unlock feature of "the" Simulation: create accurate simulation of the big bang and life.
4:28 all of those crooked chains at the bottom, do all forces they'll enact on each other on a left-right basis equate to balanced forces????? incredible how the computer could predict it
i am totally impressed. unbelievable simulations. its amazing and frightening at the same time. that much correctness is a indication of the progresses of the softwares. and thinking the possible scenarios for what could be happen, and what human kind did with the technological advancements, it is terrifying.
Impressive animations! I'm wondering why the chains are made of rubber though. Metal chains stop abruptly at the end before swaying; they don't stretch and rebound like these did.
Looks real, a sense that we are approaching something on the horizon where it's all integrated... beautiful video especially the destruction of objects!
a few papers down the line we might be able to simulate living cells. not in real time, because of the number of objects, but the physics interaction simulation are reaching a level of quick accuracy. Could you make a video about how this simulation can be adapted to the molecular level? this would be a step in an interesting direction.
Holy shit, if we can get some of these simulations, specifically this one, dumbed down a bit, rendered on the GPU, we could have some amazing physics in video game environments without big performance hits.
