Make sure to check out the paper itself here, it is a really short read and let me know if you enjoyed it! - rdcu.be/cWPfD Orig. Nature link with clickable citations: www.nature.com/articles/s41567-022-01788-5
The free link to download the paper is not working. it leads back to the Nature website, which is paywalled...it allows "read-only", not download to read for later...
Congratulation on getting your paper accepted. I love what you are going for the the community, im not in the field myself but im always amazed by the things that are possible, through your yt videos
Those really are amazing! I love these physics sim videos. Everytime you post one of these I wonder all over again how close we are to a reasonably "full" physics engine that can handle all states of matter at reasonable accuracies and interactive speeds. Between our improving graphics cards and our algorithms that improve even faster than said cards, it ought to happen eventually
Close. Probably a couple years from having all the tools, and another couple before deployment--though this is for common states (solid, liquid, gas, plasma) and not some of the weirder states (Rydberg matter, superfluids etc.). Those weirder states will probably be worked on in a different domain (purely R&D + academic)--but some papers from the last couple years suggest that neural nets may make simulation of complex quantum states many many orders of magnitude more efficient (alphafold 2 is a good example of such progress).
If you are talking about having them av in computer AAA games. I doubt that we are close. These are isolated simulations (with very limited interactivity).
@@carveratutube I'm aware. We'd still need a few hundred times speedups to a thousand times to begin to use stuff like this in actual games. Getting down from days to minutes is highly encouraging though
@@carveratutube Have you seen what NVidia have been doing in radically reducing the size of VDBs using a neural calculation? This will be great for games.
@@carveratutube These are obviously not ready for a game, yeah, but the physics calculations are getting fast enough to simulate fluid mechanics in real time within the next few years--especially given that, recently, good-quality AI approximations have run in, very nearly, real-time. Once that happens, it's a matter of integration into game engines--which will take another few. Still not that far off.
I love your work Károly. As an engineer that works with CFD and FEA tools all the time, these papers get me so excited for the future of the field. As a gamer, I can't wait for these papers to flow into real-time physics engines. What a time to be alive!
I really enjoyed this, we are SO CLOSE to being able to do this real time and make it indistinguishable to real life. What a time to be alive! My only sadness is seeing how dry the items remain, water particles that hit the dry surfaces should leave a small film of water (drying or not). Without this crucial detail, it looks very fake to me. But don't get me wrong, I am LOVING this.
I dare say these are simplified demonstration events. If someone wanted to take the personal/professional time to add surface imperfections and microscopic hygroscopic particles as part of the original texture layer the algorithm could handle that easily, it's just that these papers are coming out so fast it's more about development and usability than finished application and presentation level production detail, if you see my point.
This may sound really strange but I think one of the best ways to show this amazing work to more people would be by creating a separate channel with compilations of really unique simulations in extremely high fidelity. Even better if you can add really stimulating or even relaxing music and sound effects depending on the content. You could include all the citations and links in the description for people who wanted to dive deeper or include a short promo at the end of the vids that points to this channel with more information. I could see that kind of content doing really well. I would absolutely love watching and sharing videos like that. I’d even be interested in making that content myself if I even knew where to start with creating those simulations.
Aaand, I place, a comma after, every third, or fourth word, when, writing my own, papers, aaaand sometimes, after only, one, or two, words. Aaaand - wait for it - use the word, "aaaand", extensively, to combine statements, aaaand, to start, new sentences. Aaaand, that, is how, I keep, a steady flow, aaand, talking rhythm, aaand, create, a nice, listening experience. 😎
Love your channel and enthusiasm! Congratulations also on your publication to Nature :-) - Well done! Have been doing some simulations in 3DS Max using Pheonix FD and can't believe I'm able to also use my laptop for the simulation work. That and using real-time engines like Unreal - awesome time to be alive indeed!!
If we ever get to a point you can test the thing in real time, it would be possible to make a software where you can use "sculpting tools" to tweak the object in realtime to get to the best result.
WOW! I was very impressed stuffing around in Unreal Engine 5 with fluid sims but this is next level.. is it possible to dumb this down to mere mortals and explain what software/process is used to make these simulations?
i know its good to compare with past papers but i never know with your uploads how much of this ISNT recycled content. like im just skipping through the clips now and ive seen all of these clips before... idk...
6:28 - If this were cream I might accept it. But milk has the same consistency at water, basically, and I thought the cherries looked a bit too floaty when they went in. 6:40 - Someone did not model a thumb-tack accurately. They are not sharpened that way, nor at they that sharp. Also, thumbtacks are made with the cheapest process, so they always have clear stamping marks on the bottom, and only the top is smooth. I have not tried to float them though. I will try that tomorrow!
Thank you for sharing all of this with us. Been really enjoying your content for over a year now. It's amazing how quickly "two more papers down the line" can come! This is wonderful, and congratulations as well!
"Wow a simulation that looks like reality!" LOL Look around you man! Wow a reality the looks like a simulation said I a simulation of my Simulated Self.
So cool, thank you verry much for those quality videos. As an Engineer I would love to see some of these simulation with harsh environement (under the rain, hit by a wave, falling into water, swallowed by a tornado, taking a hit of a car at diferent speeds, heated up to melting) easily accessible.
What would you recommend someone try to learn in order to get into the field of Computer Graphics for someone whos new and nothing of the field?? Like what are some prerequisites needed to get started and successful in getting into Computer Graphics? I would love to get into this specialization.
The idea of optimizing fluid simulation with particles that scale to the detail of the interaction is not only clever in terms of saving computational horsepower, it also helps alleviate the issue of fluid simulations looking too uniform. What I mean by that is when every particle in the sim is the same scale water sims often take on a slightly jelatinous look once your eye picks up that individual particle size. I suspect this new technique will do an even better job of tricking the eye. Truly spectacular innovations on display here. I freakin love this channel 😀
2:58 This may have implications for quantum mechanical simulations, insofar as I understand, the more energetically dense a structure is (like the sun or the early universe), the more complexity and "fidelity" a structure will have when interacting with space-time, due to the mass and energy density literally distorting space-time towards itself. This is a form of time-dilation. The less energy, the fewer quantum interactions, vice versa. My particular understanding is somewhat lacking and may need some work, but regardless this paper should be seen by professional physicists studying space-time and quantum mechanics, not just computer graphics researchers. This is amazing.
For whatever reason, I’m not usually impressed by the physics graphics papers you show, but this time, your explanation of the bubble algorithm was as captivating as the footage
It's fascinating how much time, effort, resources and money go into developing such tech. But most of all, the wonderful inspiration of the scientists and researchers working on stuff like this. I think most people who benefit from this tech take it for granted. Like how all this ends up helping game development and Hollywood movies. Imagine if all of these industries invested more on research and development in science.
You can go for better than realism , “what do you mean better than realism?” How about an elephant with blue eyes? -Best programmer that has ever lived. T. D.
This is awesome! I do wonder what is the largest fluid simulation we could do in realtime with our current gpus that would still look good? Awesome video, keep up the great work!
Just had a flashback to Lagoa Multiphysics 1.0.. that was just over a decade ago... it's unbelievable how much progress has been made. The next generation of consoles will be photorealistic. The 4000 series from Nvidia will pump out a whopping 83 TFLOPS. What a time to be alive.
Congrats on your paper! The average person doesn’t know what a leaf and a paper boat floating on water looks like! I don’t know tech stuff but I know the average person. I truly on , my own level can understand and appreciate what what you and your fellow tech teams do! It’s wonderful but all I see is a future In which people wouldn’t know or care whether ‘ it’s live or is it memorex’ !
Thanks for the interesting video! Unfortunately computer graphics algorithms have yet a long way to go to match the accuracy of CFD algorithms. I remember this case where an engineer was doing a parametric design study by changing the angle of a particular part wrt the flow, and while CFD demonstrated a proportional dependence of the pressure drop wrt increasing the angle across the various cases, the computer graphics algorithm could not establish a coherent relationship and was all over the place. The problem with that is that even if the flow does look somewhat realistic, they definitely aren’t good enough for making engineering decisions. I think the existing algorithms are great for making beautiful graphics for various purposes, but significant advances are required for them to represent actual physical processes to offer an alternative for CFD yet, although various ML algorithms seem to be doing a really good job of that.
*SORRY, but did not you already use this simulation image/video with a leaf and a boat like dozens of times?* It is time to move on. All parts of this video were already shown and explained on the channel.
If aerodynamic simulations can be made cheap, we can "grade" any design and have a generative model create those designs. From embeddings to the boom of NLP, I think it's safe to say the ability to grade stuff on a grey scale is one the most important foundation for trainable AI
Each simulation seems to be focused on a some subset of phenomena. But does they work together to simulate all fluid dynamics? I was always curious about that. Single achievements are amazing, but if we cannot use them together we're still some steps behind the "ultimate solution".
All these cool videos and you say that you tried the samples, but what about sharing the samples to try it out ourselves and play with the simulations? Could be so cool...
Make sure to check out the paper itself here, it is a really short read and let me know if you enjoyed it! - rdcu.be/cWPfD
Orig. Nature link with clickable citations: www.nature.com/articles/s41567-022-01788-5
Another master piece Dr.
it's very close to the 'nature' itself :)
The free link to download the paper is not working. it leads back to the Nature website, which is paywalled...it allows "read-only", not download to read for later...
Congratulations! Keep up the wonder. You've become a worthy representative of the field.
porque usas un modulador de voz?, esta sobre editada
Congratulation on getting your paper accepted. I love what you are going for the the community, im not in the field myself but im always amazed by the things that are possible, through your yt videos
Thank you so much for being so kind! Honored to have Fellow Scholars like you in our ranks! 🙏
Those really are amazing! I love these physics sim videos. Everytime you post one of these I wonder all over again how close we are to a reasonably "full" physics engine that can handle all states of matter at reasonable accuracies and interactive speeds. Between our improving graphics cards and our algorithms that improve even faster than said cards, it ought to happen eventually
Close. Probably a couple years from having all the tools, and another couple before deployment--though this is for common states (solid, liquid, gas, plasma) and not some of the weirder states (Rydberg matter, superfluids etc.). Those weirder states will probably be worked on in a different domain (purely R&D + academic)--but some papers from the last couple years suggest that neural nets may make simulation of complex quantum states many many orders of magnitude more efficient (alphafold 2 is a good example of such progress).
If you are talking about having them av in computer AAA games. I doubt that we are close. These are isolated simulations (with very limited interactivity).
@@carveratutube I'm aware. We'd still need a few hundred times speedups to a thousand times to begin to use stuff like this in actual games. Getting down from days to minutes is highly encouraging though
@@carveratutube Have you seen what NVidia have been doing in radically reducing the size of VDBs using a neural calculation? This will be great for games.
@@carveratutube These are obviously not ready for a game, yeah, but the physics calculations are getting fast enough to simulate fluid mechanics in real time within the next few years--especially given that, recently, good-quality AI approximations have run in, very nearly, real-time. Once that happens, it's a matter of integration into game engines--which will take another few. Still not that far off.
I love your work Károly. As an engineer that works with CFD and FEA tools all the time, these papers get me so excited for the future of the field. As a gamer, I can't wait for these papers to flow into real-time physics engines. What a time to be alive!
Been looking forward to that video since I saw the community post o.O
I really enjoyed this video on Simulations. Thank you for making it!
You are very kind, thank you so much! 🙏 These simulations are truly incredible.
A.I. can also be very strong at approximating physics. It can even learn the underlying concepts itself - no need to present it the equations.
I really enjoyed this, we are SO CLOSE to being able to do this real time and make it indistinguishable to real life.
What a time to be alive!
My only sadness is seeing how dry the items remain, water particles that hit the dry surfaces should leave a small film of water (drying or not). Without this crucial detail, it looks very fake to me.
But don't get me wrong, I am LOVING this.
Don't worry shiny stuff is circa 1980's cgi 👌ua-cam.com/video/LxSeAvjmZfo/v-deo.html
I believe they left it out as they are right not focusing on only the motion of the water molecules
I believe they left it out as they are right not focusing on only the motion of the water molecules
I believe they left it out as they are right not focusing on only the motion of the water molecules
I dare say these are simplified demonstration events. If someone wanted to take the personal/professional time to add surface imperfections and microscopic hygroscopic particles as part of the original texture layer the algorithm could handle that easily, it's just that these papers are coming out so fast it's more about development and usability than finished application and presentation level production detail, if you see my point.
This may sound really strange but I think one of the best ways to show this amazing work to more people would be by creating a separate channel with compilations of really unique simulations in extremely high fidelity. Even better if you can add really stimulating or even relaxing music and sound effects depending on the content.
You could include all the citations and links in the description for people who wanted to dive deeper or include a short promo at the end of the vids that points to this channel with more information.
I could see that kind of content doing really well. I would absolutely love watching and sharing videos like that. I’d even be interested in making that content myself if I even knew where to start with creating those simulations.
that is a great idea
That's just what everyone else is doin. What's the difference then
Aaand, I place, a comma after, every third, or fourth word, when, writing my own, papers, aaaand sometimes, after only, one, or two, words. Aaaand - wait for it - use the word, "aaaand", extensively, to combine statements, aaaand, to start, new sentences. Aaaand, that, is how, I keep, a steady flow, aaand, talking rhythm, aaand, create, a nice, listening experience. 😎
"Hey Stiiiiimpy! You see dem pixels??" 👀
Robotic voice, i am scared
The voice is a simulation too... 😄😂🤣
Congrats on the paper!
I've already seen all of these on your channel.
Why are you speaking so curious? "Now ............ wait ............ we just said ............ that ............"
As someone heading to an AE career, you've got me excited about how accurate these non-cfd options are getting!
Love your channel and enthusiasm! Congratulations also on your publication to Nature :-) - Well done! Have been doing some simulations in 3DS Max using Pheonix FD and can't believe I'm able to also use my laptop for the simulation work. That and using real-time engines like Unreal - awesome time to be alive indeed!!
This man speaks at 100 periods per second.
If we ever get to a point you can test the thing in real time, it would be possible to make a software where you can use "sculpting tools" to tweak the object in realtime to get to the best result.
Wow, posted 2 hours ago, what a time to be alive !! Love that tiny boat scene.. absolutely perfect, can’t tell it was not a real filmed scene..
Congratulations and many thanks for your paper and all your efforts in making this channel!
Man, congrats on the paper. I love your channel!
Damn! I must say that as a turbine designer engineer, the idea of near real time and accurate enough simulation would be a dream come true.
WOW! I was very impressed stuffing around in Unreal Engine 5 with fluid sims but this is next level.. is it possible to dumb this down to mere mortals and explain what software/process is used to make these simulations?
Brilliant! A truly exciting time to be alive.
Congratulations Dr. Zsolnai-Fehér!! Always loved your videos, the structure and excitement is always a joy for me. Thank you!
i know its good to compare with past papers but i never know with your uploads how much of this ISNT recycled content. like im just skipping through the clips now and ive seen all of these clips before... idk...
6:28 - If this were cream I might accept it. But milk has the same consistency at water, basically, and I thought the cherries looked a bit too floaty when they went in.
6:40 - Someone did not model a thumb-tack accurately. They are not sharpened that way, nor at they that sharp. Also, thumbtacks are made with the cheapest process, so they always have clear stamping marks on the bottom, and only the top is smooth. I have not tried to float them though. I will try that tomorrow!
Congratulations. Getting a Nature Physics paper accepted is remarkable!
Wow, congratz on getting a paper in Nature! What a time to be a subscriber!
In this crazy world - it is good to see some progress being done.
Tech and scientific progress is the only progress i think.
This is amazing. Awesome. Congrats!
Thank you for sharing all of this with us. Been really enjoying your content for over a year now. It's amazing how quickly "two more papers down the line" can come! This is wonderful, and congratulations as well!
"Wow a simulation that looks like reality!" LOL Look around you man! Wow a reality the looks like a simulation said I a simulation of my Simulated Self.
I could not imagine a better ambassador for computer graphics. And thank you for making the paper freely available!
wat a time to alive!
congratulations :) it's absolutely great.
Congratulation on your publication to Nature!
Awesome work! Can't wait until this gets into Blender3D
So cool, thank you verry much for those quality videos. As an Engineer I would love to see some of these simulation with harsh environement (under the rain, hit by a wave, falling into water, swallowed by a tornado, taking a hit of a car at diferent speeds, heated up to melting) easily accessible.
That rocket gave me chills, if you ever tried to do that in blender you know
holy cow that sphere with the air bullet going trough was awesome
Congrats mate. Having something published in any Nature publication is an awesome feat. Great video, too!
What would you recommend someone try to learn in order to get into the field of Computer Graphics for someone whos new and nothing of the field?? Like what are some prerequisites needed to get started and successful in getting into Computer Graphics? I would love to get into this specialization.
The idea of optimizing fluid simulation with particles that scale to the detail of the interaction is not only clever in terms of saving computational horsepower, it also helps alleviate the issue of fluid simulations looking too uniform.
What I mean by that is when every particle in the sim is the same scale water sims often take on a slightly jelatinous look once your eye picks up that individual particle size. I suspect this new technique will do an even better job of tricking the eye. Truly spectacular innovations on display here. I freakin love this channel 😀
I waited until he said that "what a time to be alive" If he did, it means that a very important development has taken place.We think the same way.
Congratulations on your publication!
Wow! Amazing work! Could this be implemented in Blender 3D? Congrats on your paper!
this is why I Subscribed😎👌🏼
2:58 This may have implications for quantum mechanical simulations, insofar as I understand, the more energetically dense a structure is (like the sun or the early universe), the more complexity and "fidelity" a structure will have when interacting with space-time, due to the mass and energy density literally distorting space-time towards itself. This is a form of time-dilation.
The less energy, the fewer quantum interactions, vice versa.
My particular understanding is somewhat lacking and may need some work, but regardless this paper should be seen by professional physicists studying space-time and quantum mechanics, not just computer graphics researchers.
This is amazing.
So what did you get your doctorate in so you can call yourself a Dr Two minute papers?
Two Minute Papers: Food science!
Why does this guy put commas in between every few words LMAO
Nature: "Look what they need to mimic a portion of our power!"
Congratulations! 🍾
This is incredible work!
Please, use this in games or metaverse! Just call all gaming companies, they need this so much!
For whatever reason, I’m not usually impressed by the physics graphics papers you show, but this time, your explanation of the bubble algorithm was as captivating as the footage
If a more advanced civilization exists, why couldn't we be in one of these simulations ?
It's super interesting to see this kind of meta study video on a specific topic. Great perspective on the subject!
the advancements in the water simulations gives me hope that one day i'll get to play a realistic whitewater kayaking game. (preferably in VR)
It's fascinating how much time, effort, resources and money go into developing such tech. But most of all, the wonderful inspiration of the scientists and researchers working on stuff like this. I think most people who benefit from this tech take it for granted. Like how all this ends up helping game development and Hollywood movies. Imagine if all of these industries invested more on research and development in science.
You can go for better than realism , “what do you mean better than realism?” How about an elephant with blue eyes? -Best programmer that has ever lived. T. D.
This is awesome! I do wonder what is the largest fluid simulation we could do in realtime with our current gpus that would still look good? Awesome video, keep up the great work!
I love your passion for the field :)
Wow, what an amazing accomplishment! Congrats on getting your paper accepted!
Just had a flashback to Lagoa Multiphysics 1.0.. that was just over a decade ago... it's unbelievable how much progress has been made. The next generation of consoles will be photorealistic. The 4000 series from Nvidia will pump out a whopping 83 TFLOPS. What a time to be alive.
For fluid simulation we need a scale because see like this doesn't mean anything because water in a box we don't know if the box is 10cm wide or 100m
Congrats on your paper! The average person doesn’t know what a leaf and a paper boat floating on water looks like! I don’t know tech stuff but I know the average person. I truly on , my own level can understand and appreciate what what you and your fellow tech teams do! It’s wonderful but all I see is a future In which people wouldn’t know or care whether ‘ it’s live or is it memorex’ !
Thanks for the interesting video!
Unfortunately computer graphics algorithms have yet a long way to go to match the accuracy of CFD algorithms. I remember this case where an engineer was doing a parametric design study by changing the angle of a particular part wrt the flow, and while CFD demonstrated a proportional dependence of the pressure drop wrt increasing the angle across the various cases, the computer graphics algorithm could not establish a coherent relationship and was all over the place. The problem with that is that even if the flow does look somewhat realistic, they definitely aren’t good enough for making engineering decisions.
I think the existing algorithms are great for making beautiful graphics for various purposes, but significant advances are required for them to represent actual physical processes to offer an alternative for CFD yet, although various ML algorithms seem to be doing a really good job of that.
I wonder whether this approach could be used to simulate the sound of wind instruments, water flow or even a voice.
*SORRY, but did not you already use this simulation image/video with a leaf and a boat like dozens of times?*
It is time to move on. All parts of this video were already shown and explained on the channel.
0:39 the Anti-R world? Does that run on arm or x86?
If aerodynamic simulations can be made cheap, we can "grade" any design and have a generative model create those designs.
From embeddings to the boom of NLP, I think it's safe to say the ability to grade stuff on a grey scale is one the most important foundation for trainable AI
lots of repeated content from past videos. hard to even tell what is the new content.
Each simulation seems to be focused on a some subset of phenomena. But does they work together to simulate all fluid dynamics? I was always curious about that. Single achievements are amazing, but if we cannot use them together we're still some steps behind the "ultimate solution".
some pretty incredible demos
I just wish that few papers down the line we might one day get all this in realtime and get implemented in video games.
Please make a video on how to get stable diffusion working with a Lambda cloud account!
It' s amazing all this kind of investigation.
Having recently seen cherries floating in water, the edges around those looked oddly static. But had I not, im not sure i'd have seen the difference.
I wish some day to see you at the Internet Festival in Pisa. That would affect so many students here!
two minute papers with his 11 minute videos. more to enjoy.
Ok that's it. Seeing how quickly this tech advanced has convinced me we live in a simulation lol.
Thank god that u recovered from that simulation.
Cool. Thanks for sharing.
Congratulaions Karoly! Pzublication in Nature is a serious career achievement. You earned it buddy!
Congratulations on getting your paper accepted into such a prestigious journal. Looking forward to one day cutting hours off my Blender fluid sims.
Congrats on the Nature publication 🙌
Hold on to your papers!
Great video and topic.
What a time to be alive 😁
it's always inspiring hearing a scientist talking about something they deeply care
I love it every time he says "What a time to be alive"
Damn, two minutes going over 10 minutes now
"Devilishly difficult task"😂😂
This is the channel that makes me most happy
All these cool videos and you say that you tried the samples, but what about sharing the samples to try it out ourselves and play with the simulations?
Could be so cool...
Day by day we are closer to MATRIX!
The world of computing is getting more exciting every other day
Since we finally have 100tf+ graphics cards coming out next year I wonder if a scaled down version is possible in a real-time graphics engine.
Boy, that water simulation technology is going to make video games much more immersive.
I love this kind of research it really motivates me and my soul.
This is what happens when you all use equations in math and physics to a simulation.
It is funny how it is still immediately recognizable when some real photo / video is shown.