Are we hearing this from the perspective of a microphone within the liquid? Or is it transferring through the water / out the container as well? It sounds very convincing, however somewhat "underwater" if that makes sense.
I think it's just modeling the viscocity of water at a different temperature than we usually hear, see this video: th-cam.com/video/Ri_4dDvcZeM/w-d-xo.html
@@anselmschueler No the video has nothing to do with this video. They have not computed water temperature and have just got averages. You did not answer his question, which the answer to is actually what he initially said, the virtual microphone is placed in the water opposed to outside
It sounds like there are still some high frequency noises missing. The bubbles' freq also seems to be a bit too constant and almost musical. I feel like they should have more imperfect frequency shifts, but maybe it is just me. Great job you guys!
I remember thinking about doing this in the late 90's but was far too ignorant of the relevant mathematics to be able to build a model at the time. Incredible that college students are doing work like this routinely now.
Does this also take into consideration the temperature of the liquid? Since the viscosity of water changes according to temperature, the sound changes as well.
Not by much. You'll also note that the math (The Helmholtz problem) doesn't reference the viscosity anywhere, but in the derivation there's a reference to the navier stokes, which includes a viscous diffusivity term. I'm sure that there's some inputs there, but, to be frank bubbles are bubbles...
All the sounds would have been easier to follow if you showed the FFTs and zoomed in close enough to see where the sounds are originating. Looking at drops from across the room and only hearing the sound does not give the experience you gained by looking at the microscopic behavior. You considered the tiny surfaces of the drops and bubbles, but only show macroscopics (particular at 3:00 and extensions). This is probably good, but you lost a lot of the immediacy and power but putting the camera too far away, and the time scale at human speeds. When the actual time scales are connected to the frequencies of the sounds and their time periods.
+Banana Warrior We used Gerris (gfs.sourceforge.net/wiki/index.php/Main_Page) to simulate the fluid and BEM++ (www.bempp.org/) to solve the boundary integral equations.
+Banana Warrior Maxwell Render (www.maxwellrender.com/), mainly because it supports nested dielectrics (support.nextlimit.com/display/mxdocsv3/Nested+dielectrics), which was the easiest way to handle the three interfaces correctly.
Timothy Langlois I understood, like, 40% of what you have said but thanks anyway man Now i have to figure out if this is even possible in houdini (Since i'm really new at the VFX stuff and never heard about those programs D:)
basically, what he was saying is that having several transparent things inside each other is hard and only few renderers can handle it, Maxwell being one of them. - The air, glass and water all have different optical properties.
Wow. But I think I would *never* notice that the sound of water pouring into some tank in an animated movie/game was pre-recorded in another environment rather than synthesized/simulated with a physically correct model. But thanks for the physics and techniques!
Interesting. For the shown examples, the frequency and popping extension seem to have little to no notable effect, with only the micro bubbles being significant, and full transfer is either not significantly different from no transfer, or seems worse for the dam break, with proxy transfer being more clear while the others sound muffled.
What is the software name? Did you develope the code by yourself? And in which programming environment? Would it be possibile to reprocude a similar calculation in Matlab? And most of all, what are the hardware resources you used to make these simulations?
I wasn't listening with headphones but I'm guessing the lossy sound compression of TH-cam is responsible for the 'attack of the sine waves' effect I'm hearing. I note there is an uncompressed audio file available here www.cs.cornell.edu/projects/Sound/bubbles/ . I'll have to download it and have a listen.
iLikeTheUDK I believe mp3 and similar compression systems use FFT like analysis (probably DCT or wavelet) to cut down the number of frequencies needed to fool our hearing. By throwing away unnoticed frequencies they cut down the data needed to represent sounds. This fails when the input is wide band noise, lots of frequencies all over the spectrum. So instead of water splashing you hear discrete sine waves all over the frequency range.
Looking at the python code for this paper, you guys are using 1497 m/s for the speed of sound. However, that's the speed of sound in seawater, but your paper is [presumably] for fresh water (c ~ 1435 m/s) ? I'm guessing it won't have much of an effect, but it was enough to make me stop and double check when trying to replicate your results.
It will slightly change the damping of the bubble oscillators, but I doubt the effect will be noticeable. For most of the derivation we're assuming the liquid is incompressible (implies the speed of sound is infinite).
@@TimothyLanglois Thanks for the unexpected, but definitely welcome reply. Most of the research out there seems to be on simulating the acoustics of single bubbles at a time, but our application needs to simulate the acoustics of millions (or at least hundreds) of bubbles (from for example, ocean wave splash transients) in real-time. Do you know of any research out there that covers the statistics of simulating acoustics for large numbers of bubbles (that isn't just wide-sense stationary spectra)?
I always wondered when they'd start to simulate sound with physics inside the computer... maybe one day we'll be able to model characters and have them talk without voice actors :)
You should make this into a VST Plugin! As a music producer, I would use it a lot for adding background accents and color to my songs. Right now, If I want a water sound I have to synthesize it manually which is much harder and less realistic than this video.
Making this a vst plugin would be very cpu intensive because: 1. You need to simulate the liquid first (which will be time consuming) 2. You also neex to simulate the sound (which is also time consuming as the render time is slow) So basically just record sounds from the real life. It's faster
![[SIGGRAPH 2018] Toward Wave-based Sound Synthesis for Computer Animation](http://i.ytimg.com/vi/su6z9snjU-U/mqdefault.jpg)
![[SIGGRAPH 2018] Toward Wave-based Sound Synthesis for Computer Animation](/img/tr.png)
![[SIGGRAPH 2018] A Multi-Scale Model for Simulating Liquid-Fabric Interactions](http://i.ytimg.com/vi/xvyGpBKevLM/mqdefault.jpg)
Well.
Now I know what dropping water shaped as armadillo sounds like. :D
Impressive.
Perunapiirakka `
Are we hearing this from the perspective of a microphone within the liquid? Or is it transferring through the water / out the container as well? It sounds very convincing, however somewhat "underwater" if that makes sense.
I think it's just modeling the viscocity of water at a different temperature than we usually hear, see this video: th-cam.com/video/Ri_4dDvcZeM/w-d-xo.html
@@anselmschueler No the video has nothing to do with this video. They have not computed water temperature and have just got averages. You did not answer his question, which the answer to is actually what he initially said, the virtual microphone is placed in the water opposed to outside
just reverb it lmao
It sounds like there are still some high frequency noises missing. The bubbles' freq also seems to be a bit too constant and almost musical. I feel like they should have more imperfect frequency shifts, but maybe it is just me. Great job you guys!
Swoonpoon you are right, it sounds like you have one ear under water and the other above.
What's their obsession with those armadillos?
graphics.stanford.edu/data/3Dscanrep/ (they're free, save time and make for good fun.)
Lost it at those. :) Thanks for the explanation and source Lucas!
The tank with flowing water sounds exactly like it should if you record the sound with piezo fixed to the side of the tank. Good job, very impressive!
Now they just need to add the effects from the resonating air column above the tanks...
I remember thinking about doing this in the late 90's but was far too ignorant of the relevant mathematics to be able to build a model at the time. Incredible that college students are doing work like this routinely now.
If by "college students doing this routinely" you mean "PhDs publishing research in fluid simulations" sure
Soon this might be part of the state of the art of sound design.
Does this also take into consideration the temperature of the liquid? Since the viscosity of water changes according to temperature, the sound changes as well.
Not by much. You'll also note that the math (The Helmholtz problem) doesn't reference the viscosity anywhere, but in the derivation there's a reference to the navier stokes, which includes a viscous diffusivity term. I'm sure that there's some inputs there, but, to be frank bubbles are bubbles...
@@patrickscholl7919 th-cam.com/video/Ri_4dDvcZeM/w-d-xo.html
All the sounds would have been easier to follow if you showed the FFTs and zoomed in close enough to see where the sounds are originating. Looking at drops from across the room and only hearing the sound does not give the experience you gained by looking at the microscopic behavior. You considered the tiny surfaces of the drops and bubbles, but only show macroscopics (particular at 3:00 and extensions). This is probably good, but you lost a lot of the immediacy and power but putting the camera too far away, and the time scale at human speeds. When the actual time scales are connected to the frequencies of the sounds and their time periods.
Great work! Is it possible to calculate the liquid-solid interactions with these method?
Which programs do they use to... simulate this stuff?
+Banana Warrior We used Gerris (gfs.sourceforge.net/wiki/index.php/Main_Page) to simulate the fluid and BEM++ (www.bempp.org/) to solve the boundary integral equations.
Timothy Langlois Thanks man
What'd you use to render it though?
+Banana Warrior Maxwell Render (www.maxwellrender.com/), mainly because it supports nested dielectrics (support.nextlimit.com/display/mxdocsv3/Nested+dielectrics), which was the easiest way to handle the three interfaces correctly.
Timothy Langlois I understood, like, 40% of what you have said but thanks anyway man
Now i have to figure out if this is even possible in houdini (Since i'm really new at the VFX stuff and never heard about those programs D:)
basically, what he was saying is that having several transparent things inside each other is hard and only few renderers can handle it, Maxwell being one of them. - The air, glass and water all have different optical properties.
Wow. But I think I would *never* notice that the sound of water pouring into some tank in an animated movie/game was pre-recorded in another environment rather than synthesized/simulated with a physically correct model. But thanks for the physics and techniques!
Interesting. For the shown examples, the frequency and popping extension seem to have little to no notable effect, with only the micro bubbles being significant, and full transfer is either not significantly different from no transfer, or seems worse for the dam break, with proxy transfer being more clear while the others sound muffled.
Is it just me, or does the proxy transfer sound better than the full transfer?
1:40 That's E# and F
When can I port this into minecraft?
I LOVE this. this is amazing! Well done guys
What is the software name? Did you develope the code by yourself? And in which programming environment?
Would it be possibile to reprocude a similar calculation in Matlab?
And most of all, what are the hardware resources you used to make these simulations?
I wasn't listening with headphones but I'm guessing the lossy sound compression of TH-cam is responsible for the 'attack of the sine waves' effect I'm hearing. I note there is an uncompressed audio file available here www.cs.cornell.edu/projects/Sound/bubbles/ . I'll have to download it and have a listen.
"attack of the sine waves" I like that name.
iLikeTheUDK I believe mp3 and similar compression systems use FFT like analysis (probably DCT or wavelet) to cut down the number of frequencies needed to fool our hearing. By throwing away unnoticed frequencies they cut down the data needed to represent sounds. This fails when the input is wide band noise, lots of frequencies all over the spectrum. So instead of water splashing you hear discrete sine waves all over the frequency range.
what is program name for this simulation ? visual c ,open gl function?
Excellent work! I would expect this functionality to be in Blender (3D suite) one day, hopefully.
the future of ASMR
This is really really really really cool
Awesome work!
Looking at the python code for this paper, you guys are using 1497 m/s for the speed of sound.
However, that's the speed of sound in seawater, but your paper is [presumably] for fresh water (c ~ 1435 m/s) ?
I'm guessing it won't have much of an effect, but it was enough to make me stop and double check when trying to replicate your results.
It will slightly change the damping of the bubble oscillators, but I doubt the effect will be noticeable. For most of the derivation we're assuming the liquid is incompressible (implies the speed of sound is infinite).
@@TimothyLanglois Thanks for the unexpected, but definitely welcome reply.
Most of the research out there seems to be on simulating the acoustics of single bubbles at a time, but our application needs to simulate the acoustics of millions (or at least hundreds) of bubbles (from for example, ocean wave splash transients) in real-time.
Do you know of any research out there that covers the statistics of simulating acoustics for large numbers of bubbles (that isn't just wide-sense stationary spectra)?
That armadillo is a must-have to every Siggraph simulation technical paper. Duh, everyone knows that lol xD
W=(ma^2)/2 How is this derived?
v is not velocity in this case so v dot is not acceleration. v is the volume offset (previous slide).
@@gregmcdonnell4652 Ok thanks, I should be reading instead on the paper. Maybe it was explained there.
Amazing. My mind is blown.
I wonder how many people at Siggraph had to go to the bathroom after that presentation
I always wondered when they'd start to simulate sound with physics inside the computer... maybe one day we'll be able to model characters and have them talk without voice actors :)
The matrix: simulates everything
These guys: yes.
This is a good start but not all bubbles terminate on the surface so quickly.
Looks good
So, proxy transfer is the best.
No. It doesnt work on all models and variations
Hey can you guys make this into a VST so musicians can use it? I would buy the shit out of that
I had no idea that they were working on audio sims too
You should make this into a VST Plugin! As a music producer, I would use it a lot for adding background accents and color to my songs. Right now, If I want a water sound I have to synthesize it manually which is much harder and less realistic than this video.
Making this a vst plugin would be very cpu intensive because:
1. You need to simulate the liquid first (which will be time consuming)
2. You also neex to simulate the sound (which is also time consuming as the render time is slow)
So basically just record sounds from the real life. It's faster
Thats amazing!
Wow. Impressive
amazing
This is the physics of water in video games of 2050
Wow... just wow
The matrix has you.
Now you just need to add splashes on the walls with drip marks. 😀
neat!
Why do you always use bunnies and armadillos as test subjects? Just a small question coming from 2 years later.
Complicated form, a good way to show resolution of finite element models and a good way to show the depth of simulation solution spaces
And 10 more months later : graphics.stanford.edu/data/3Dscanrep/
This is freaking me out
This sounds like pouring water from underwater.
your work will never be used but good job
I thought this was real.
Matrix, step closer
i want drink water
im thirsty.
Ахренеть.
Ъ