Beautiful Fluid Simulations...In Just 40 Seconds! 🤯
HTML-код
- Опубликовано: 28 май 2021
- ❤️ Check out Weights & Biases and sign up for a free demo here: wandb.com/papers
❤️ Their mentioned post is available here: wandb.ai/wandb/getting-starte...
📝 The paper "Wave Curves: Simulating Lagrangian water waves on dynamically deforming surfaces" is available here:
visualcomputing.ist.ac.at/publ...
🙏 We would like to thank our generous Patreon supporters who make Two Minute Papers possible:
Aleksandr Mashrabov, Alex Haro, Alex Serban, Andrew Melnychuk, Angelos Evripiotis, Benji Rabhan, Bryan Learn, Christian Ahlin, Eric Haddad, Eric Martel, Gordon Child, Haris Husic, Ivo Galic, Jace O'Brien, Javier Bustamante, John Le, Jonas, Kenneth Davis, Lorin Atzberger, Lukas Biewald, Matthew Allen Fisher, Mark Oates, Michael Albrecht, Nikhil Velpanur, Owen Campbell-Moore, Owen Skarpness, Ramsey Elbasheer, Robin Graham, Steef, Taras Bobrovytsky, Thomas Krcmar, Torsten Reil, Tybie Fitzhugh, Ueli Gallizzi.
If you wish to appear here or pick up other perks, click here: / twominutepapers
Károly Zsolnai-Fehér's links:
Instagram: / twominutepapers
Twitter: / twominutepapers
Web: cg.tuwien.ac.at/~zsolnai/ Наука
Q: What does one fluid simulation say when meeting another fluid simulation?
A: Nothing, they just wave!
Boooo!
Q: What do we call people who always post lame jokes in comment sections ?
A: Dumbass
@@thetruth7
Q: Can you taste it? Really?
A: Yes, its salty and toxic. Its the truth.
Wow, you /stole/ a bad joke? Boooo!
@@jayxi5021 Yes. Very toxic. Edited. Confused. My bowyee. Jayxi.
Whoa so smart to change the particle size based on how much is going on in a specific region!
It's essentially an R-tree
The first time you see it, it's amazing. The concept isn't that hard to come up with but implementing it efficiently is probably very hard. (I haven't done that part myself yet)
the idea of partitioning that way has roots all the way back to octtrees / r-trees
h-refinement!
I was thinking the same, that's a brillant idea!
Oh snap. Oh my Papers.
They implemented this on a CPU! And not even a new one!
True, it's not always easy to implement these on a GPU, but sooner than later *it will be done* . And it will be blazing fast.
What A Time To Be Alive!!! 😍
what it is not in a gpu???
so this can easy run in realtime on newest cards of optimised correctly.
Damn, didn't even realize that. Couldn't this be close to realtime on GPU then? Pretty amazing. If they had water like that in a game, pretty sure I could just sit there watching it for hours.
from what I understood of the video, you need a pre-existing simulation made with particles to be able to add waves in those 20-ish seconds, using a CPU. if i got this correctly, it means that adding those cool waves could be done in real time on a GPU, but the whole simulation wouldn't be able to run in real time.
@@TheZenytram yeah any parallelizable thing over thousands of cores MASSIVELY increases speed
I want to see a simulation that lets water flow past objects that aren’t programmed to be hydrophobic, like how water flowing over a rock will make the rock shiny where it still have a thin layer of water on it even when the water has moved passed it
Yea that would be cool. Abs not technically so hard.
Check out wetmaps in blender. I have a video demonstrating that effect in fluid sim I did in a sink. It's not a very good animation but I assume that's what you're referring to.
This is beyond the scope of the papers because it is more related to materials and shaders than how particles behave.
In fact it is super simple compared to the things shown here, they just didn't bother calculating collisions to change shader properties.
This paper covers fluid-material simulations.
Is Simulating Wet Papers Possible?
ruclips.net/video/_4fL4jnC8xQ/видео.html
You should have mentioned that their code was implemented for CPU and run on 8 core CPU. WIth 64 core Thread Ripper, or GPUs, it will be almost real time, at least for lower quality version.
Woah
Ah,.. that is rather important info. But I'll be really impressed when it can add spray as well. The example was hinting at such, but it wasn't visualised.
This is actually crazy
Only if it can be implemented on a GPU. Some algorithms require a CPU.
This is even more valuable for future games. You can have a realtime fluid simulation on a nice creak and have it look good too! Maybe not quite there yet without many optimizations but there's hope for a actually physically accurate rivers and creaks reacting to stuff you throw in them. What a time to be alive!
Some games already use similar yet more simplified approach since some years back
“Creek”
WTX 8090 Ti
Water Tracing cores
To my eyes, all the examples shown here were too exaggerated, making things look unnatural. Being conservative with the number of wave curves added will give very subtle enhancement adding to realism while decreasing computation time.
Not all of them, but a few certainly didn't feel like water
I believe that's mostly because there's no froth.
@@micaelgarcia1576 Right, some of them felt like actual strings were put in the water.
I thought 200k looked more natural than 800k.
This is just the simulation. Put it into a nicer mask, and add some bubbles. That doesn't seem hard.
it is cool but it looks that the water lacks of tension surface
Sorface tansion!!
tension surf-ace
I think it's opposite it looks to me as if the water has more surface tension than normal
@@aman_singh__ More surface-tension -> smoother surface in the face of flow-interference.
Less surface tension -> more ripples and less smooth and coherent surface.
OMG!! U knocked this out of the ball park!!! the W&B is amazing!! & the personal account will truly ''help'' me 2 get going!! thank U so much 4 this Dr. !!
Still waiting for all these amazing new papers to reach blender :(
I found something similar called - Flow maps, but its not that good
what is the usual translation time for that?
It's already there way before this "paper" =) blendermarket.com/products/vortex
implent it
@@stanislav.shelest This vortex plugin isn't a simulation, it mostly a bunch of artistic postprocessing procedural shaders, with a lot of controls. It doesn't produce realistic stuff, but is fast, is more of visually nice and desired, which is of course great. But not the same as the new results presented in the video.
As someone who lives by a lot of flowing water, these wave curves add so much more realism to the simulation. So good
The surface looks unrealistic with all those waves at once
Yea liquids don't get waves jeeeeezeee
It needs surface tension
It's because of the flat shader, it would look more realistic with a water shader.
It's not the simulation, shading is the issue. Our skin is very rough and bumpy, but we can't really see it that much because of subsurface scattering. The same is the case here.
@@johnschwalb Not like that they don't
I love how this channel can bring such esoteric and niche topics to the public in an easy to understand way. It's also given me a new found appreciation for computer graphics - I'll never watch a pixar film the same way...
I love your videos so much, they inspire me a lot...
Added to reading list 👍
Great work by authors👏
Thank you sir for inspiring and keeping students like us motivated to study hard 😊
What a time to be alive!!
You could use a function of wave curve “point density” (where lots of them are crossing) to generate wash/white water pretty easily (increased “point density” == increased opacity of the rendered area) and that might look even more realistic!?
impressive as always. can't wait to use this
Your channel deserve a lot of love
This one is very impressive, looks great!
Hello, I love your videos, you bring attention to some really cool computer graphics stuff. :)
You are very kind. Thank you! 🙏
@@TwoMinutePapers hey doc ...
I couldn't believe this ..but look ...REALTIME WATER PHYSICS IN A GAME !!!and even an old one ...the source engine can do THIS :
ruclips.net/video/PzKoTPtbtZE/видео.html
COULDN'T BELIEVE IT !!
I love this channel, you should win awards for all the work you do!
This video got me wondering, are there some of these papers that can be applied in real time on consumer GPUs? For gaming and such, but with a much higher level of quality?
My jaw dropped when he unblurred that left half....those micro-ripples are that missing fluid sim feature that I didn't even know I was missing😭 Absolutely incredible!
The new normal for fluid Sims.
I don't know why but this flashed me moore than any other paper I saw so far. Maybe it's because I lover so much.
Just use waves to simulate waves 😎😂
genius 😂😂
Next video: "We used balls to simulate balls"
*@Two Minute Papers: I hope you see this message.* This may sound kind of dumb, but I want to tell you that your channel has become my favorite spot on RUclips and I look forward to watching your videos each week. Your channel has become a source of happiness in my life. I just wanted to tell you that 😊
Just love it when you go "wow..." 😀
Using curves! wow thats smart, never thought of it
This water looks amazing!!!
amazing work!
Truly beautiful 🌊
‼️ Can't wait for all these things to be added into softwares like Blender for normal consumers to use. This channel is awesome
(copy pasted because it's been asked by others)
I actually made an addon a few months ago after being inspired by the original video of this paper. The addon uses a different method to generate the details, but in my (biased) opinion that method looks slightly better because it's not as wrinkly.
My paper (not published or reviewed): drive.google.com/file/d/1_aU2wntd20n6La9kiwJHvTFI4dOOZQ8A/view?usp=sharing
The addon: blendermarket.com/products/vortex
It looks kind of unrealistic in some situations, but I could see this being used to great effect for skin rendering.
This is just what I was thinking, looks almost more like wrinkles in leather/skin to me rather than waves on a liquid surface.
It looks weird to me. Like stringy water. It would be perfect for rendering the skin on cooked milk or pudding though
Delicious slow-motion jelly as always when someone wants to simulate water.
The best way to watch this channel is every couple months go back and be a-maaaazed
We will finally have videogames with insanely beautiful water?! 😍😁 Love it. Thanks for this, Two Minutes Papers!
It's crazy how these simulations just get better and better :0
Did you really think it was an improvement on the reference sim (the state of the art?)? It looked kind of fake to me, it'd be interesting to see how it holds up to reality. Very pretty though, rich in detail for sure. But stylized.
It looks like how the surface of a flowing molten metal makes ripples as the surface viscosity changes cooled by the air like a thin film
True
Now our water can have wrinkles
Yeah that was my thought too. I don't particularly like how it looks. I think the long almost perfect wrinkles are what sets off the strange look.
@@ElectricalSwift i think they are just a bit too strong
absolutely amazing
wow WOW
look at those wave curves - i want a deep dive on wave curves!!
Can't wait to see these sorts of simulations playing out in video games!
Oh wow this looks just like the Multigrid methods I'll be working on, impressive
It puts a smile on my face every time he says Dr. before his name.
So beautiful!
Omg what I would DO to hire you to do fluid simulation ads for my skincare brand!!!
Now this is *_truly_* the first strand-type fluid simulation.
I thought the title was going to say "40 minutes" and I was like "alright alright pretty cool" then I opened the video and saw it said "seconds" and I was like "BRO THAT'S DOPE"
what a coincidence. i was literally just looking up whether unreal engine can create light refraction effect under water. the way random waves refract light different exactly like this, and creates light affects on surfaces underwater...That would be a good next realism step after this one for water sims. Using this new tiny wave data to project light patterns under the water.
Quick question. What happens to all these papers? I mean, I see so many amazing stuff on your channel but I really wish to use some of them in Blender.
Is there a way to implement them in Blender. And if so, why isn't it been done? These stuff are to amazing to leave aside.
Thank you sir.
Water simulations always sort of looked like loose jelly, this looks like real water and put into the right environment with some small additional details such as wetting rocks and sand and adding white caps, no one could tell the difference. I love this so much!!!
Should have told us to hold onto our papers at 0:55. The visual difference between the two is amazing. Later, when it is made translucent, you probably could have fooled me by saying it was real. What a time to be alive!
Anyone else just have no clue what the actual technical stuff is but find this amazing anyway?
These simulations are incredible.
That is really cool and look super realistic, i guess I'll have to wait until it gets added to blender 😂
HOLY SHIT... thats ridiculously good
in your videos, when you talk about the time it takes to process a simulation, what kind of computer are we talking about?
a super computer or a personal computer?
MINDBLOWING
What algorithms are used to render the water in the first place? Is it all screen space or are solid geometries of triangles being constructed somehow?
The water looks like a winkly sphinx cat after you touch it's back.
Ah, yes, that's exactly the analogy I was looking for, thanks.
As a programmer and 3D artist by far one of my favorite videos..check this > 3:21 there's a air bubble at the right?? That looks AMAZING!
When is it coming out for Blender?
Great paper, the number of waves looks like dependent of the viscosity of the fluid or some physical quantitty, for water looks greate for 100k waves, but for higher values is more like something non newtonian fluid by the number folds
Impressive fluid sims! What software? How would one implement in Blender?
Anyone else listen to this just for this guys nice voice or is that just me?
Hey Dr., I know its not the same as this paper but have you ever seen the program AmberGen? Its a Particle Simulator Software that runs in real-time for Smoke, Fire, and other Cool FX. Its kinda of limited to what it can do but for what it is, its pretty cool It's got lots of parameter sliders to control the particles behavior in many ways. Every slider is able to be keyframed to an animation timeline in the software so the user can create some really cool FX by keyframing and changing the parameter they want down the timeline. I dont think it can import/export tracking data with to be used with 3D compositing, so its mainly for exporting to use in video compositing software like Adobe After FX, Nuke, Davinci Resolve, etc.) or Photo Based Software to add some flare to Artwork.
Hi! Which software can do this? Thanks!
These wave seem more artistic than natural. But great work. I think they are still usable in certain way in natural simulation
love this
Amazing!
I'm not sure what it is but it looks like the small waves don't interfere with each other, making it look the water contains many long physical strings that somehow don't get tangled.
It's a cool effect though and it does make the water look more high resolution, just in an unusual way.
Mmm, crispy details!
well they look nice, but do they have any physical significance or is CFD still the way to go for that
Hmm, I wonder if this sort of surface modifier could be applied to smoke? Would be a neat way to add vortices
This looks like adding, exactly as you said "High Frequency info" into the low frequency data. We perceive reality in FFT spectra as I realized from working with light and images of wavelengths reflecting off of water. It's some amazing stuff that you are going in depth with this knowledge with perhaps, not knowing this hidden fact of perception and physics?
Basically, I realized by looking at water flow the other day that reflections, due to their dim light, show "corse" data, i.e. the low freq FFT data within the water. The main reflected light on the surface of the water shows the high spectra FFT that we see and associate with the water. It's pretty complicated stuff, but I theorize that if we (you amazing computer scientists) can come up with a way to decompose the almost 3d nature of the FFT spectra of the waves reflecting off of the 2D surface of the water, then you could theoretically input Just the Light data from a trainer to learn how to model water in a simulator in a completely new way. I propose that this FFT method may be faster or at very least reveal new insights into how to capture / add different "coarseness" / "fine detail" FFT spectra to our fluid simulations.
Either way, there clearly is hidden data within the course dataset from what I have observed and studied just with images of light and FFT decomposition. Im studying/ working on how FFT spectra will redefine all of our future studies into Physical phenomenon as a multi layer potential way to view reality in all of its forms. Quite puzzling stuff. Reach out to me if you would like to know more. I've decomposed light interference to see the "filament" like structure of light from a simple purple and blue image. It's quite freaky stuff honestly to be able to see the actual "fundamental fabric / nature" of reality so easily with FFT decomposition.
I wonder is that possible to apply these to Nvidia Flow realtime liquid simulation, if I be able to export as Alembic or USD (not sure if it contains geometry animations).... Its be just great in gamedev/archviz....
Excellent...
Animated Dream's shirt: look what they need to mimic a fraction of our power
It's important to note that the linear scaling is opposed to something like exponential scaling, where twice the amount of curves could take more than double, usually more
Yay new video
Wow that is cool! Free surface simulations are difficult.
imagine this in real time that would be soo awesome
Is this just for aesthetic purposes or can it be used for scientific and engineering purposes? Thanks for the content, I especially love the fluid simulation videos.
At this phase it makes it look as though there is a thin skin on the surface of the liquid. But I'm excited to see what this evolves into. The approach of post-processing simulations sounds very promising.
oh, this is going to be great for video games !
Woah, now that's cool
Wave/particle hybridization would be good. Would be funny if Heisenbergian factors result.
this type of simulation can support in unity built in render pipeline ?
Does it mean 1 frame in 10 seconds (w/ 100,000 wave curves)? Or/and can we use it in video games?
I believe the technique used is amazing but, can it decide if a fluid doesn't need those wave lines? I mean, lets say a body of water gets turbulence by an object that goes into it, after that initial perturbance: can it decide the rate at which those wave lines disappear until there is none? Or it has artifacts (an unnecessary amount wave lines)?
I see rivers in a new way.
ps. linear time :O love it
What GPU do you use exactly?
how can i use it in my blender simulation?
I hope some time in the future this can be simulated in real time.
Where i can use it?
What kind of rendering engine does these researchers use to render these kind of simulations? Because all the material and such needs be setup. Or do they do it completely from scratch, so just like rendering in the very early days?
We need a simulation for simulated water droplets and bubbles. That is what gives the liveliness of water. It makes the water look like their is always something alive under it or interacting with it. And not just ripples. This is some very intriguing papers, though.
It is interesting that we're simulating the wave/particle duality to get realistic fluid surfaces.
What if we apply this to steam and smoke?
What about the relasitic wetting of objects