I've worked as a fisherman for a decade. A lot of that time was spent watching the sea and thinking 'there must be a logarithmic mixture to reproduce this effect'. Here comes Acerola to crown himself again.
16:32 once I was trying to make the surface ripples of water running in a direction like a stream, and one of my attempts had small ripples that were just moving along the surface without any disturbance and I thought it was unrealistic and added some noise to it, then I saw water running down the street in the rain and it had super static unchanging ripples running along the surface, I felt really stupid
I swear to god staring at a thing in reality, questioning whether or not I would consider it realistic if it was presented in a video game is too relatable you explained a few concepts in this video that I took like, 2 years to figure out on my own, damn you! (but thanks for sharing it for others) now I can finally point people to a specific good video when they ask for shader knowledge
On shadertoy there's a shader called "Seascape". Even with the source code right in front of me, I've never been able to understand it. This was a phenomenal explanation and I feel like I actually understand it now.
Nobody does it like you. Explained well, entertaining, thorough, explained performance pitfalls and how to solve them, etc. Your work is a goldmine. Thank you!
@@ai_outline i mean you gotta sacrifice one or two braincells to fry to create this magnificent juicy water. Not as Avatar 2 level of water but convincing splashy water.
I spent all day making structs do what I want them to do in my midi sequencer. This is what I needed to unwind. Hell yeah, that was some comfy overlap between audio and graphics. So many topics that are extremely familiar to someone who has spent more time with synthesizers than with code. Ultimately it all comes back to the Fourier transform when doing things involving sums of sines. Also one thing used in audio design to reduce noticeable beating (i.e. the audio analogy of tiling) is to vary each oscillator independently as a function of time. Plenty of ways to solve that issue but that's the one that immediately comes to mind for me.
I am constantly amazed at your ability to make complex topics easy to understand and fun to watch. As an aspiring game dev and current 3D artist, thank you.
I'm a physics student going into grad school, and I'm planning on working on real time fluid simulations! It's fascinating seeing it from a nonphysical, more approximate perspective. Of course Fourier series would make their appearance! The domain warping aspect is very fascinating! I would never have thought of that, but it works perfectly. It also seems to naturally result in the low regions being flat and low detail, and the high regions being spiky and high detail. That's so cool! It's also fascinating learning about all this optics stuff. That's the one bit of physics I never got. I suppose if I want to work on games (a hobby of mine) I'll have to learn it some day...
Acerola, you're a lifesaver. You just saved the surfing mechanic I was thinking about scrapping because until now my waves looked like an unyielding herd of gelatinous migratory hillocks. still gotta work out a more performant way to make the big waves that crash and go tubular etc., but at least now it feels like I'm on the ocean, and not lost in some sort of undulating mound dimension.
this method I still think looks yucky for bigger waves, it looks really nice for slower water/stillwater distortion on reflections. Next vid will go over the logic behind the nice realistic big waves though!
Awesome video, I learned a lot! Still two minor corrections/annotations: 1) frequency is not 2/wavelength, but 1/wavelength. Probably that mistake somehow slipped in with the graph at 1:46. You can see there that you are not showing sin(x) but sin(PI*x). The natural wavelength of a sinewave is 2*PI, also known as a radian. Basically this is an angle-measurement that is based upon the arclength (= traversing the surface) of a circle with the radius of one. 2) the Blinn-Phong model is NOT in any way the basis for PBR. Blinn-Phong is an approximation on the basis on artistic criteria, calculated purely for verisimilitude, not realism. PBR models on the other hand approach the calculations on the basis of physical values. For example a light in Blinn-Phong can be brighter going out than it goes in since it is added once in the diffuse and a second time in the specular breaking energy conservation. A PBR model on the other hand only distributes the total amount of light into reflection and absorbtion, so it can never be brighter than it goes in. Oh and I didn't expect you to be this young, with the quality of your videos and complexity of the topics discussed I expected you to be an industry veteran for at least a few years!
water is one of my favorite topics when it comes to computer graphics because there's so many ways to go about implementing it. my personal favorite video game water is Mario Galaxy, because it feels really stylized and unique, without resorting to solid colors and repetitive textures like a lot of other examples for stylized water. The toxic waste in Portal 2 also looks really nice, despite basically being brown sludge.
Something to mention with Sea of Thieves, you actually can see the repetition in the water from high viewing angles i.e. climbing a mountain. It is just very expertly hidden behind all of their other fancy work.
yeah this is cause the fft approach is too expensive to compute in the shader so it's instead computed into a texture that tiles, otherwise there wouldn't be repetition.
@@Acerola_t That makes a fair amount of sense. It's still something I'm really interested in learning as well. Most of my stuff is done using textures too to be fair, though lately I've been baking heightmap and other data out of Houdini water sims to get an actual simulation as the baseline and then work with everything outside of that. Have you also seen their big talk about their cloud rendering system? It's quite an approach they use, they gave a whole breakdown at GDC 2019 called Sea of Thieves: Tech Art and Shader Development. I think it might be worth looking at as an approach as well for your other videos coming up, since you mentioned you were getting into cloud rendering as a subject soon too (iirc in your CS2 smoke grenade video)
Yes! I have been hoping for a good detailed video about FFT and iFFT realistic water. I would still like one on that spooky partial differential equation one too, but I know that it's got hands
Dude, I'd be lying if I said I learned or comprehended much of what you just did, I'll be sixty next year and my short-term memory is, what was I saying? What I can say truthfully is I was able to follow along with you the whole video, you didn't lose me, my eyes didn't glaze over, and my mind didn't drift off. That's good, you have a knack for this! I wish you had been around when I was in high school, I'd have learned a lot more math and enjoyed math a lot more instead of dreading it out of fear I'd become a CPA..
You explain those topics better than my professor, it's hard for me to understand certain equations without having an intuitive grasp about it. Your visualization really helps!
It's cool to see a breakdown of how stuff usually works, but it's much rarer to see any content explaining 'better' options. Excited for the next video :)
Acerola and Sebastian Lague collab when?! I also really like Sebastian's work. I remember obsessing over ant simulations about 2 years ago when I saw his ant/slime simulation videos! lol
I started my college degree of computer science in the beginning of this year, and even though my primary objective is not to become a graphics designer and also because i am still a noob when it comes to programing and everything, your videos still are always really amazing, informative and fun to watch. Ngl, you are even better to explain somethings than my teachers are lol, hope you grow even more here on TH-cam you deserve it
Love the videos, mad props. One thing I do wish though is that we didn't get a 101 to graphics programming (what a normal is, what diffuse/specular are) in each video. Perhaps just refer people back to an earlier one? I know it's a difficult balance to strike to adjust it correctly for the target audience. Keep up the awesome videos!
Ever since I started watching your videos, I've been looking at real objects and thinking about the way it could be modelled digitally. Water has been the primary object of my attention, so I was ecstatic to see an Acerola water shader video! Thank you so much and I'm looking forward to more water shading stuff in the future :D
Only problem with the sound addition at 21:09 is of surf sounds. When on the ocean, you don't hear waves doing anything but hitting the side of your boat sometimes. Being on a beach sounds like the surf, but being a mile out to sea sounds a lot more quite since there is almost no sound (from water) of waves hitting things. You do hear the wind on when out on the sea, depending on the height of things on your ship.
your honestly one of the game dev people that I like to watch I find it more better of someone explaining me how it's done than someone just tell me how to make it and just coping the code without no knowledge of how it's done, and you're more entertaining to watch than a 2 hour video of someone explaining it like you video are more faster but not to fast to were I don't understand and can't follow along
Nice video! This helped a lot in my understanding of lighting, other explanations I saw were really dry and didn't explain the math very well. You should also cover distortion textures! By abusing distortion and multiple scrolling textures, you can make most visible tiling go away. There's a great video by a channel named Jasper about how they used this for water in Mario Galaxy 2.
First and foremost: Beautiful results, excellent in-depth info, very entertaining and funny. Acerola continues to be one of the best graphics programming edutainment channels I know on YT. Looking forward to the follow-up video, but no pressure. We should be thankful to get these, so I am now considering signing up for Patreon, if there isn't a better (more direct) way to support these accessible, relevant and honestly simply good videos. On topic: A few years ago I created an ocean system for a game that was featuring a lot of water, both visually and mechanically. As pointed out, that Gerstner waves aren't that easy to control and that is a very good point and the reason I chose to not use them. If you need a lot of game design control over water, I found that simple and small look-up textures did the trick. Artists set up simple animation curves in the editor along with additional parameters (calm vs. stormy water looks very different, so you want different contributions from different frequency bands), which were then baked on start-up and evaluated in simulation steps on the CPU and GPU respectively. So, artists could shape waves to look like Gerstner or any other type of non-overlapping wave. They could also have multiple peaks at different heights etc. Amplitudes were controllable locally, as it was just another runtime simulation parameter, so it was easier to control the chaotic patterns when the situation demanded it. You don't want players to be frustrated, because they got unlucky with an extreme trough and couldn't reach a safe spot in time. The solution was very fast on last generation hardware and its synchronous but independent implementation meant that no data beyond basic control parameters needed to be transferred between processors. The GPU computed a height map (there was more than just waves that shaped the water) for rendering and the CPU simply computed samples on demand when queried by gameplay systems, so no height map was required (we also had orders of magnitude fewer queries than the sample count required to compute such a map).
Great video, just a small suggestion: when writing named functions in LaTeX (e.g. sin, cos, etc.), use "\[function name](x)" instead of just typing the name (e.g. "\sin(x)" instead of "sin(x)"). That way it won't be italized as in 14:25 (the math jumpscare), and would instead be typeset as a normal text.
yeah that is the literal one latex visual in which i forgot to use the \text for it, if you look at every single other one in the vid it looks correct lol
I really hope you get around to the more advanced wave methods. The result at the end of this video looks great but sea of thieves' waves are something else.
sea of thieves waves take up half the screen 80% of the time. You bet your ass they have teams of dedicated, nda-locked programmers and artists, who are as smart and hard-working as, if not more than, a single sleep-deprived youtuber, focused solely on the water.
@@sumdude5172 They did a siggraph talk in 2018 going over it. The talk was called "The Technical Art of Sea of Thieves," it's on youtube. I'm not a developer so maybe they left out some important details, but it seems pretty complete from what I can tell.
the physics of how objects interact with the water and disrupt the usual wave patterns is something I always wanted to learn. Any modern game with seafaring in it would feel incomplete without it
Every other Ace video I go like that "Oh" gif, starring Chi McBride No better way to convey how this feller here makes me understand stuff that hadn't really sunk in
Having watched a lot of your videos back to back, it's pretty apparent how good you've gotten with pacing and delivering just enough information to explain concepts without getting lost in the details. Awesome work as always, looking forward to the next one! You did raise a good question that I've never seen a satisfying answer for: why DO games include a bunch of post-processing effects that a lot of people always turn off? Are those people in the minority? Are the effects often just not done well? Is it mostly for consoles/lower-end hardware where lower framerates need more fancy effects to compensate?
Speaking of sine waves, here's a neat graphics curve you can use: `sin(t * pi) ^ 0.15`. It quickly fades in and out at the outer edges near t=0 and t=1 respectively. Although I've seen weird numeric errors where when t=1, the output is not exactly 0.
I’ve worked with gerstner waves quite a bit and I’ve only ever noticed issues with waves curling in on themselves if you add a wave to the mix with parameters higher than the one that came before it. I’d be interested in seeing you talk more about gerstner waves or explain a bit more why you don’t like them!
tbh the real reason I dont like them is it heavily complicates the partial derivatives due to the horizontal displacement and I found that the e^sin(x) waves looked great anyways for less effort (mentally and computationally)
@@Acerola_t I can agree with that. They certainly add way more complexity than needed for some water use cases. I already find derivatives annoying and adding the non linear, multi dimensional nature of gerstner waves to the equation is a bit much lol Much love can’t wait to see what you upload next!
Maybe I'm being dumb because I don't know graphics or gamedev at all (I just find your videos extremely entertaining and interesting), but around 12:26 shouldn't the camera vector be inverted (pointing the other way) for the math to work? If I add te vectors as drawn I get a horizontal vector as H, and therefore would get something close to 0 when doing the dot product with the vertical normal vector shown. Maybe instead of adding I should substract, I think that would work too. In any case, great video! Keep the content coming!
I do like this implementation of water, it looks quite natural and great at the same time, the approach I generally go for is just to take two noise textures and use procedural sampling so that it doesn't tile but I might start using a similar approach to this in the future. also i got a bit confused because I completely forgot about the water community post and was looking on your GitHub post processing effects for different tonemappers and saw the lens flares so yea lmao, still a great video and very informative
After the opening of this video and seeing all the monogatari figures, I can finally understand why your title screens always remind me of that show! It was inspired by it!
Sine wave and all periodic stuff is extremely useful in a wide variety of applications, even if they are just a bit related to math. I always laugh when I see people make the classic "Another day without cosine" joke.
14:00 - You also need to sample the collision points at least 3 times to get an accurate surface function for the CPU. It was honestly really simple once I got hang of it.
I liked the fact that you chose something else than gerstner waves, which is commonly accepted for being the go-to for waves simulations, you explained why and proved that you could still have good looking water without that much complexity…
21:37 Is there any possibility for somehow using primes in the wave functions to decrease the frequency of tiling or is that sort of technique only really useful for integer based patterns?
The repetition doesn't come from the number of divisors, it comes from sin() and the size of the waves. To reduce the repetition you would need to add more, larger waves that constructively and destructively interfere with all of the other waves over a larger area, making the repetition more spread out.
@@filiformis if I take like, sine(x) + sine( sqrt(2) * x) (where sqrt(2) is an arbitrarily chosen irrational number) , shouldn’t that, not repeat? Of course, rational numbers can approximate an irrational one arbitrarily closely, so I guess it should get closer and closer to repeating, but I don’t get why that would apply for distances that aren’t *really* big? Is this something I just need to actually do the calculations for?
@@drdca8263 Irrational numbers don't exist in computers. Everything gets rounded off at some point. And regardless, we're not trying to get a wave that doesn't repeat. We're trying to create a wave that doesn't look like it repeats. A wave that has a discernible pattern that slightly changes over millions of units technically "doesn't repeat", but looks like it does. A wave that repeats once every millions of units does repeat, but looks like it doesn't. But the code is available. You can give the sin waves irrational numbers in whatever way you think will make them not repeat and see what happens.
@@filiformis I guess I should have explicitly mentioned that I was aware that on computer it wouldn’t actually be irrational. Still, if the continued fraction for a rational number has many terms, my impression is that in some contexts it acts more “as if it were irrational” than other rational numbers with short such expansions, and so I thought that perhaps that could apply here, and so, a close approximation to an irrational number would result in taking a long time to repeat. I also acknowledged the possibility of approximate repetition. Perhaps I should have explicitly acknowledged that approximate repetition could look bad for the same reason. I was expressing that I expected that the distance before things would approximately repeat would be so large to be not a problem, but seeing as this is apparently not the case, I am led to ask: why does this approximately repeat (I.e. look visually similar enough to be an issue) frequently enough to be an issue? If for some epsilon > 0, and some D, L > 0 and some x_0, then if for all x in [x_0,x_0+L], |f(x+D) - f(x)| < epsilon, then say that f epsilon-approximately repeats an interval of length L at a distance D. Would this capture the type of repetition that we are concerned about here? if we fix x_0 and L, and let f_r(x) := sine(x) + sine(r x) I wonder, for each epsilon, what is the smallest D such that f_r epsilon-approximately repeats an interval of length L starting at x_0, at distance D? And, how does this vary with r? f_r(x+D) - f_r(x) = (sine(x+D) - sine(x)) + (sine(r (x+D)) - sine(r x)) .. Hm, for sine(x)-sine(y) to be within some distance, then, x-y should I think be within some distance of an integer multiple of 2pi? Need to make this precise... hm, I’ll think on this
The algorithm seems to have picked on me talking about caustics and reccomended me "VFX Artist Explains the HARDEST Visual Effect to Make" from corridor crew. I wouldn't mind at all a third video in this series about caustics to tight the knot.
21:40 Wouldn't the tiling pattern be hidden, at least partially, if some sort of gradient noise was added to to the sum of sines ? I assume there's a good reason why it is not done, and I'd be curious to know what it is. Too expensive ? Maybe it makes the waves look unnatural ?
21:33 If you lift the camera up, you could do something like LevelOfDetail in the texture world. E.g. decrease the sinus waves and mix them with other new waves, which comes with the new content om the screen edges where new content shows up. With this extra code you can perfectly finish your video about water animation.
first off. amazing video as always, goes without saying since your quality is literally unmatched not only in the graphics rendering space but across youtube as a whole secondly, your hair looks great
Timberborn has some pretty interesting fluid sims. It’s a semi minecraftian game in that the world and landscape is made of blocks, but the water actually flows and fills the world, as well as having levels at up to 0.01 of a block height, having waves, and even being used up by plants in the area.
I absolutely love how you explain things. Would it be possible for you to create a video about other, more modern light diffuse, specular, and reflective calculations? I'd love to see them get the same breakdowns as your explanations for sines here. The CookTerrance model for example is incredibly confusing and I feel like I'm parroting it when I'm writing it rather than understanding it.
This is just what I needed for my RTS, only don't need the render, just the waves. It's an aether map, some gameplay stuff, so I'll need to figure out how to get back the data to the cpu eventually. Thank you very mucho!!
Screen space reflections still beat cube maps but are less consistent, I like valves idea of using both screen space reflections and cube maps, mixing methods. If an improvement leads to predictable edge cases, plug the gap with old faithful.
I've been learning opengl recently and i wrote my own code for a wave, but it was just one sine wave, and the vertices were...cubes. And i was computing the position in the cpu. Never thought i should do it in the gpu, that's brilliant. Thanks aceroler
0:59 but there are real time fluid simulations look at liquigen. It runs on the gpu i think and work in realtime with fairly good performance depending on the scale of the scene.
when i say stuff like this I mean in the context of games. you have to have the rest of the game too, you cannot dedicate 105% of your resources to a single small asset.
@@Acerola_t yea that's true but I can definitely see it or something similar coming to games in a couple years. Like something similar to portal, not with a whole sea but for puzzles where you use water physics to solve them. That would definitely be very cool and probably not that computationally intensive to run.
So I’ve been watching you for a while and I’ve always loved watching these videos despite me barely understanding anything. This was the first time I watched an acerola vid and felt like I actually understood most of it, so much so that it motivated me to download unity and learn shader coding to replicate this water effect. I actually managed to get a nice looking sum of sines with lambertian diffuse and blinn-phong specular! I’ve never touched graphics programming or even a game engine before and somehow this video single handedly dragged me past the barrier. But I’ve kinda been learning exclusively shader programming and avoiding learning Unity itself, apart from like, “Shader Lab” *shudders* Now it seems I actually have to learn Unity to make a script that automatically generate waves with decreasing amplitudes and wavelengths instead of entering those parameters manually
This makes me super happy to hear! If you didn't you should definitely try reading the references in the description to sorta absorb the theory through your applied experience, that's what works best for me. There's also plenty of ways to extend the project by either doing what you said and trying to create a sort of automated weather system that can blend between different sums, or trying to layer multiple sums of differing scales to reduce the tiling dilemma, or experimenting with different waveforms, or improving the water shader itself by implementing a more sophisticated physically based shader or maybe trying to optimally stylize instead, or perhaps you wanna try and simulate buoyancy and float some objects in the water (it's pretty easy!). It's a super exciting space to work in!!!
Get a free 30 day trial and 20% off an annual plan at brilliant.org/acerola #ad
Hope you like the video!!
nice
have you considered Tessendorf's model for Fluid Sims?
you said it was free!
Nice haircut 👍
swagilicious man
this isnt just how to make water, this is a crash course in graphics programming
I've worked as a fisherman for a decade. A lot of that time was spent watching the sea and thinking 'there must be a logarithmic mixture to reproduce this effect'. Here comes Acerola to crown himself again.
The man himself has come to bring us another well explained video about complex topics I will forget as soon as I try to do it myself
this is the most accurate comment ever
16:32 once I was trying to make the surface ripples of water running in a direction like a stream, and one of my attempts had small ripples that were just moving along the surface without any disturbance and I thought it was unrealistic and added some noise to it, then I saw water running down the street in the rain and it had super static unchanging ripples running along the surface, I felt really stupid
Laminar flow babyyyyyy
@@Litl_Skitl that isn't what laminar flow is but ok
@@Solutrait is
@@simonfrancis110 captain dissolution has done an entire video on laminar flow, watch that before you say anything else
@@Solutra Sorry I thought you meant ripples that stayed perfectly in place. I need to read more carefully...
BABE WAKE UP-
Poor babe. Who can catch a nap in this house?
It's a new acerola video!!
babe, babe please why aren't you waking up
It was 3 years ago, @@naiknaik8812 it wasn't your fault...
@@kkTeazLook at me, son. It's not your fault.
I swear to god
staring at a thing in reality, questioning whether or not I would consider it realistic if it was presented in a video game is too relatable
you explained a few concepts in this video that I took like, 2 years to figure out on my own, damn you! (but thanks for sharing it for others)
now I can finally point people to a specific good video when they ask for shader knowledge
i literally do this all the time lol. when he mentioned it in the video, i let out a sigh of relief that im not the only one who does this 😅
On shadertoy there's a shader called "Seascape". Even with the source code right in front of me, I've never been able to understand it. This was a phenomenal explanation and I feel like I actually understand it now.
GPU Gems that are explained simply could be a whole series. You should do more of these.
Nobody does it like you. Explained well, entertaining, thorough, explained performance pitfalls and how to solve them, etc. Your work is a goldmine. Thank you!
basically becoming game dev also means becoming mathematician for that juicy wet water
I would say a computer scientist hahaha
@@ai_outline i mean you gotta sacrifice one or two braincells to fry to create this magnificent juicy water. Not as Avatar 2 level of water but convincing splashy water.
@eveningcommenter6312 i would love that on vr
I spent all day making structs do what I want them to do in my midi sequencer. This is what I needed to unwind.
Hell yeah, that was some comfy overlap between audio and graphics. So many topics that are extremely familiar to someone who has spent more time with synthesizers than with code. Ultimately it all comes back to the Fourier transform when doing things involving sums of sines. Also one thing used in audio design to reduce noticeable beating (i.e. the audio analogy of tiling) is to vary each oscillator independently as a function of time. Plenty of ways to solve that issue but that's the one that immediately comes to mind for me.
it's all waves all the way down!
(also that sounds like a good idea)
I am constantly amazed at your ability to make complex topics easy to understand and fun to watch. As an aspiring game dev and current 3D artist, thank you.
I'm looking forward to the next video. The waves in Sea of Thieves look so impressive, they even have physics!
I'm a physics student going into grad school, and I'm planning on working on real time fluid simulations! It's fascinating seeing it from a nonphysical, more approximate perspective. Of course Fourier series would make their appearance!
The domain warping aspect is very fascinating! I would never have thought of that, but it works perfectly. It also seems to naturally result in the low regions being flat and low detail, and the high regions being spiky and high detail. That's so cool!
It's also fascinating learning about all this optics stuff. That's the one bit of physics I never got. I suppose if I want to work on games (a hobby of mine) I'll have to learn it some day...
21:26 I'm sorry to say this, but with this rendering time you brought up a great Polish meme
Acerola, you're a lifesaver. You just saved the surfing mechanic I was thinking about scrapping because until now my waves looked like an unyielding herd of gelatinous migratory hillocks. still gotta work out a more performant way to make the big waves that crash and go tubular etc., but at least now it feels like I'm on the ocean, and not lost in some sort of undulating mound dimension.
this method I still think looks yucky for bigger waves, it looks really nice for slower water/stillwater distortion on reflections.
Next vid will go over the logic behind the nice realistic big waves though!
Awesome video, I learned a lot!
Still two minor corrections/annotations:
1) frequency is not 2/wavelength, but 1/wavelength. Probably that mistake somehow slipped in with the graph at 1:46. You can see there that you are not showing sin(x) but sin(PI*x). The natural wavelength of a sinewave is 2*PI, also known as a radian. Basically this is an angle-measurement that is based upon the arclength (= traversing the surface) of a circle with the radius of one.
2) the Blinn-Phong model is NOT in any way the basis for PBR. Blinn-Phong is an approximation on the basis on artistic criteria, calculated purely for verisimilitude, not realism. PBR models on the other hand approach the calculations on the basis of physical values. For example a light in Blinn-Phong can be brighter going out than it goes in since it is added once in the diffuse and a second time in the specular breaking energy conservation. A PBR model on the other hand only distributes the total amount of light into reflection and absorbtion, so it can never be brighter than it goes in.
Oh and I didn't expect you to be this young, with the quality of your videos and complexity of the topics discussed I expected you to be an industry veteran for at least a few years!
1) gpu gems is wrong then idk
2) I added an annotation saying blinn phong is no longer considered pbr by today's standards
water is one of my favorite topics when it comes to computer graphics because there's so many ways to go about implementing it. my personal favorite video game water is Mario Galaxy, because it feels really stylized and unique, without resorting to solid colors and repetitive textures like a lot of other examples for stylized water. The toxic waste in Portal 2 also looks really nice, despite basically being brown sludge.
Something to mention with Sea of Thieves, you actually can see the repetition in the water from high viewing angles i.e. climbing a mountain. It is just very expertly hidden behind all of their other fancy work.
yeah this is cause the fft approach is too expensive to compute in the shader so it's instead computed into a texture that tiles, otherwise there wouldn't be repetition.
@@Acerola_t That makes a fair amount of sense. It's still something I'm really interested in learning as well. Most of my stuff is done using textures too to be fair, though lately I've been baking heightmap and other data out of Houdini water sims to get an actual simulation as the baseline and then work with everything outside of that.
Have you also seen their big talk about their cloud rendering system? It's quite an approach they use, they gave a whole breakdown at GDC 2019 called Sea of Thieves: Tech Art and Shader Development.
I think it might be worth looking at as an approach as well for your other videos coming up, since you mentioned you were getting into cloud rendering as a subject soon too (iirc in your CS2 smoke grenade video)
Yes! I have been hoping for a good detailed video about FFT and iFFT realistic water. I would still like one on that spooky partial differential equation one too, but I know that it's got hands
“wow frass nal!” was far more funny than it should have been
Dude, I'd be lying if I said I learned or comprehended much of what you just did, I'll be sixty next year and my short-term memory is, what was I saying? What I can say truthfully is I was able to follow along with you the whole video, you didn't lose me, my eyes didn't glaze over, and my mind didn't drift off. That's good, you have a knack for this! I wish you had been around when I was in high school, I'd have learned a lot more math and enjoyed math a lot more instead of dreading it out of fear I'd become a CPA..
You explain those topics better than my professor, it's hard for me to understand certain equations without having an intuitive grasp about it. Your visualization really helps!
Trigonometry and fluidity are two topics I would have never have guessed would apply in graphics together. Now, it makes sense!
Trove of invaluable tricks as always! Thanks!
Thank you so much!!
It's cool to see a breakdown of how stuff usually works, but it's much rarer to see any content explaining 'better' options. Excited for the next video :)
You and Sebastian Lague have made me realise I want to be a graphics programmer, and I thank you for that.
THIS!
This field is so interesting!
Good luck!! You can do it!
Acerola and Sebastian Lague collab when?!
I also really like Sebastian's work. I remember obsessing over ant simulations about 2 years ago when I saw his ant/slime simulation videos! lol
I started my college degree of computer science in the beginning of this year, and even though my primary objective is not to become a graphics designer and also because i am still a noob when it comes to programing and everything, your videos still are always really amazing, informative and fun to watch.
Ngl, you are even better to explain somethings than my teachers are lol, hope you grow even more here on TH-cam you deserve it
Have fun!
also graphics designers make logos not video game graphics, important distinction!
@@Acerola_t noted 👍🏻
It's such a pleasure to watch an ugly looking plane of sine waves slowly but surely becoming a masterpiece
Love the videos, mad props. One thing I do wish though is that we didn't get a 101 to graphics programming (what a normal is, what diffuse/specular are) in each video. Perhaps just refer people back to an earlier one?
I know it's a difficult balance to strike to adjust it correctly for the target audience. Keep up the awesome videos!
I'd rather each of my vids be a complete package
Totally agree about screen space reflections, they're like ants: once you notice them you see them everywhere and it totally ruins immersion.
Ever since I started watching your videos, I've been looking at real objects and thinking about the way it could be modelled digitally. Water has been the primary object of my attention, so I was ecstatic to see an Acerola water shader video! Thank you so much and I'm looking forward to more water shading stuff in the future :D
Only problem with the sound addition at 21:09 is of surf sounds. When on the ocean, you don't hear waves doing anything but hitting the side of your boat sometimes. Being on a beach sounds like the surf, but being a mile out to sea sounds a lot more quite since there is almost no sound (from water) of waves hitting things. You do hear the wind on when out on the sea, depending on the height of things on your ship.
Holy shit, the amount of information is incredible, also i love how you actually manage to keep my adhs-ruined attention span
your honestly one of the game dev people that I like to watch I find it more better of someone explaining me how it's done than someone just tell me how to make it and just coping the code without no knowledge of how it's done, and you're more entertaining to watch than a 2 hour video of someone explaining it like you video are more faster but not to fast to were I don't understand and can't follow along
20:13 LITERALLY ME AFTER I LEARNED ABOUT HOW POWERFUL FRESNEL IS
Nice video! This helped a lot in my understanding of lighting, other explanations I saw were really dry and didn't explain the math very well.
You should also cover distortion textures! By abusing distortion and multiple scrolling textures, you can make most visible tiling go away. There's a great video by a channel named Jasper about how they used this for water in Mario Galaxy 2.
Watching this at 0.25x speed to offset the goober who watched at 2x speed. Gotta internalize every detail.
19:20 I cant believe Acerola forgot to mention the 4th forbidden reflection technique of just using a 2nd camera to render the reflections
basically the cubemap approach
First and foremost: Beautiful results, excellent in-depth info, very entertaining and funny. Acerola continues to be one of the best graphics programming edutainment channels I know on YT. Looking forward to the follow-up video, but no pressure. We should be thankful to get these, so I am now considering signing up for Patreon, if there isn't a better (more direct) way to support these accessible, relevant and honestly simply good videos.
On topic: A few years ago I created an ocean system for a game that was featuring a lot of water, both visually and mechanically. As pointed out, that Gerstner waves aren't that easy to control and that is a very good point and the reason I chose to not use them. If you need a lot of game design control over water, I found that simple and small look-up textures did the trick. Artists set up simple animation curves in the editor along with additional parameters (calm vs. stormy water looks very different, so you want different contributions from different frequency bands), which were then baked on start-up and evaluated in simulation steps on the CPU and GPU respectively. So, artists could shape waves to look like Gerstner or any other type of non-overlapping wave. They could also have multiple peaks at different heights etc.
Amplitudes were controllable locally, as it was just another runtime simulation parameter, so it was easier to control the chaotic patterns when the situation demanded it. You don't want players to be frustrated, because they got unlucky with an extreme trough and couldn't reach a safe spot in time. The solution was very fast on last generation hardware and its synchronous but independent implementation meant that no data beyond basic control parameters needed to be transferred between processors. The GPU computed a height map (there was more than just waves that shaped the water) for rendering and the CPU simply computed samples on demand when queried by gameplay systems, so no height map was required (we also had orders of magnitude fewer queries than the sample count required to compute such a map).
I don't know if you know how good these videos are, but if you do then well deserved! Coming from someone who's been programming for 35 years now.
Great video, just a small suggestion: when writing named functions in LaTeX (e.g. sin, cos, etc.), use "\[function name](x)" instead of just typing the name (e.g. "\sin(x)" instead of "sin(x)"). That way it won't be italized as in 14:25 (the math jumpscare), and would instead be typeset as a normal text.
yeah that is the literal one latex visual in which i forgot to use the \text for it, if you look at every single other one in the vid it looks correct lol
@@Acerola_t True! Sorry, didn't notice.
Missed opportunity to call the title "Turning sine into water"
I really hope you get around to the more advanced wave methods. The result at the end of this video looks great but sea of thieves' waves are something else.
sea of thieves waves take up half the screen 80% of the time. You bet your ass they have teams of dedicated, nda-locked programmers and artists, who are as smart and hard-working as, if not more than, a single sleep-deprived youtuber, focused solely on the water.
@@sumdude5172 And this prevents a high level exploration of some of the methods how?
give me 2 weeks idk
@@Acerola_t 2 months it is, got it 👍
@@sumdude5172 They did a siggraph talk in 2018 going over it. The talk was called "The Technical Art of Sea of Thieves," it's on youtube. I'm not a developer so maybe they left out some important details, but it seems pretty complete from what I can tell.
the physics of how objects interact with the water and disrupt the usual wave patterns is something I always wanted to learn.
Any modern game with seafaring in it would feel incomplete without it
Every other Ace video I go like that "Oh" gif, starring Chi McBride
No better way to convey how this feller here makes me understand stuff that hadn't really sunk in
Having watched a lot of your videos back to back, it's pretty apparent how good you've gotten with pacing and delivering just enough information to explain concepts without getting lost in the details. Awesome work as always, looking forward to the next one!
You did raise a good question that I've never seen a satisfying answer for: why DO games include a bunch of post-processing effects that a lot of people always turn off? Are those people in the minority? Are the effects often just not done well? Is it mostly for consoles/lower-end hardware where lower framerates need more fancy effects to compensate?
gamers only notice when they look bad and don't have the knowledge to identify the effect when it looks good
Speaking of sine waves, here's a neat graphics curve you can use: `sin(t * pi) ^ 0.15`. It quickly fades in and out at the outer edges near t=0 and t=1 respectively. Although I've seen weird numeric errors where when t=1, the output is not exactly 0.
I have been waiting for this video to come out! YEEEESSSSSS Thank you!!!!
IM SO HAPPY you have so many monogatari figures. bro thats my favorite anime of all time. bro just HAD to show them all off
I’ve worked with gerstner waves quite a bit and I’ve only ever noticed issues with waves curling in on themselves if you add a wave to the mix with parameters higher than the one that came before it.
I’d be interested in seeing you talk more about gerstner waves or explain a bit more why you don’t like them!
tbh the real reason I dont like them is it heavily complicates the partial derivatives due to the horizontal displacement and I found that the e^sin(x) waves looked great anyways for less effort (mentally and computationally)
@@Acerola_t I can agree with that. They certainly add way more complexity than needed for some water use cases. I already find derivatives annoying and adding the non linear, multi dimensional nature of gerstner waves to the equation is a bit much lol
Much love can’t wait to see what you upload next!
Maybe I'm being dumb because I don't know graphics or gamedev at all (I just find your videos extremely entertaining and interesting), but around 12:26 shouldn't the camera vector be inverted (pointing the other way) for the math to work? If I add te vectors as drawn I get a horizontal vector as H, and therefore would get something close to 0 when doing the dot product with the vertical normal vector shown. Maybe instead of adding I should substract, I think that would work too.
In any case, great video! Keep the content coming!
you are correct yes, my visualization was wrong.
This dude is actually the goat! What a generous man to be spend his free time doing this for us
I do like this implementation of water, it looks quite natural and great at the same time, the approach I generally go for is just to take two noise textures and use procedural sampling so that it doesn't tile but I might start using a similar approach to this in the future. also i got a bit confused because I completely forgot about the water community post and was looking on your GitHub post processing effects for different tonemappers and saw the lens flares so yea lmao, still a great video and very informative
I tried to implement a decent lens flare for the video but it kinda just looked like shit so I cut it
@@Acerola_t yea i was a bit confused lmao, i thought it might have been like an early version or something, still a cool idea though
As a UI programmer your videos have helped me understand more about graphics than any other. Well done chap
9:59 shoutout the Professor
Man, wish I could make videos this entertaining. Is it just a special gift you have? Really cool
After the opening of this video and seeing all the monogatari figures, I can finally understand why your title screens always remind me of that show! It was inspired by it!
Sine wave and all periodic stuff is extremely useful in a wide variety of applications, even if they are just a bit related to math. I always laugh when I see people make the classic "Another day without cosine" joke.
Thanks
14:00 - You also need to sample the collision points at least 3 times to get an accurate surface function for the CPU. It was honestly really simple once I got hang of it.
I really appreciate the quality of content that you produce! I can't remember having ever had such good, yet fun explanation on any topic!
I liked the fact that you chose something else than gerstner waves, which is commonly accepted for being the go-to for waves simulations, you explained why and proved that you could still have good looking water without that much complexity…
21:37 Is there any possibility for somehow using primes in the wave functions to decrease the frequency of tiling or is that sort of technique only really useful for integer based patterns?
The repetition doesn't come from the number of divisors, it comes from sin() and the size of the waves. To reduce the repetition you would need to add more, larger waves that constructively and destructively interfere with all of the other waves over a larger area, making the repetition more spread out.
@@filiformis if I take like, sine(x) + sine( sqrt(2) * x) (where sqrt(2) is an arbitrarily chosen irrational number) , shouldn’t that, not repeat?
Of course, rational numbers can approximate an irrational one arbitrarily closely, so I guess it should get closer and closer to repeating,
but I don’t get why that would apply for distances that aren’t *really* big?
Is this something I just need to actually do the calculations for?
@@drdca8263 Irrational numbers don't exist in computers. Everything gets rounded off at some point. And regardless, we're not trying to get a wave that doesn't repeat. We're trying to create a wave that doesn't look like it repeats. A wave that has a discernible pattern that slightly changes over millions of units technically "doesn't repeat", but looks like it does. A wave that repeats once every millions of units does repeat, but looks like it doesn't.
But the code is available. You can give the sin waves irrational numbers in whatever way you think will make them not repeat and see what happens.
@@filiformis I guess I should have explicitly mentioned that I was aware that on computer it wouldn’t actually be irrational. Still, if the continued fraction for a rational number has many terms, my impression is that in some contexts it acts more “as if it were irrational” than other rational numbers with short such expansions, and so I thought that perhaps that could apply here, and so, a close approximation to an irrational number would result in taking a long time to repeat.
I also acknowledged the possibility of approximate repetition. Perhaps I should have explicitly acknowledged that approximate repetition could look bad for the same reason.
I was expressing that I expected that the distance before things would approximately repeat would be so large to be not a problem, but seeing as this is apparently not the case, I am led to ask: why does this approximately repeat (I.e. look visually similar enough to be an issue) frequently enough to be an issue?
If for some epsilon > 0, and some D, L > 0 and some x_0, then if for all x in [x_0,x_0+L], |f(x+D) - f(x)| < epsilon, then say that f epsilon-approximately repeats an interval of length L at a distance D.
Would this capture the type of repetition that we are concerned about here?
if we fix x_0 and L, and let f_r(x) := sine(x) + sine(r x)
I wonder, for each epsilon, what is the smallest D such that f_r epsilon-approximately repeats an interval of length L starting at x_0, at distance D?
And, how does this vary with r?
f_r(x+D) - f_r(x) = (sine(x+D) - sine(x)) + (sine(r (x+D)) - sine(r x)) ..
Hm, for sine(x)-sine(y) to be within some distance, then, x-y should I think be within some distance of an integer multiple of 2pi? Need to make this precise... hm, I’ll think on this
Im just sitting here nodding at the screen as if i have any idea of whats going on, loved the video btw
The replay stats on various parts of this video are gonna go crazy and I love it
Saw Punpun on your shelf and immediately hit the subscribe button. I already know you have good taste so the videos have to be just as good.
I finally found a use for all the knowledge obtained from calculus classes: being able to understand Acerola's videos
The algorithm seems to have picked on me talking about caustics and reccomended me "VFX Artist Explains the HARDEST Visual Effect to Make" from corridor crew. I wouldn't mind at all a third video in this series about caustics to tight the knot.
21:40 Wouldn't the tiling pattern be hidden, at least partially, if some sort of gradient noise was added to to the sum of sines ? I assume there's a good reason why it is not done, and I'd be curious to know what it is. Too expensive ? Maybe it makes the waves look unnatural ?
BABE WAKE UP NEW ACEROLA VIDEO JUST DROPPED
21:33 If you lift the camera up, you could do something like LevelOfDetail in the texture world. E.g. decrease the sinus waves and mix them with other new waves, which comes with the new content om the screen edges where new content shows up. With this extra code you can perfectly finish your video about water animation.
Not that I wasn't already enjoying the video, but your Phase/Faze joke at 2:27 earned the like from me 😅😂
Love the video, great topic. Would love to see more water and PBR rendering topics
Nice haircut, just as fresh as your water. Those specular highlights.
Is your intro and style of editting inspired by Studio SHAFT or the monogatari series
first off. amazing video as always, goes without saying since your quality is literally unmatched not only in the graphics rendering space but across youtube as a whole
secondly, your hair looks great
The madlad casually dropping his entire Monogatari figure collection, based as hell.
Just wanted to say that I adore your videos. They're wonderful.
16:50 thank God it isn't just me that thinks about this! :D
Timberborn has some pretty interesting fluid sims. It’s a semi minecraftian game in that the world and landscape is made of blocks, but the water actually flows and fills the world, as well as having levels at up to 0.01 of a block height, having waves, and even being used up by plants in the area.
I absolutely love how you explain things. Would it be possible for you to create a video about other, more modern light diffuse, specular, and reflective calculations? I'd love to see them get the same breakdowns as your explanations for sines here. The CookTerrance model for example is incredibly confusing and I feel like I'm parroting it when I'm writing it rather than understanding it.
yeah i did that the video is called how photorealistic and stylized graphics are the same
Wow, expertly breakdown the details to the most basic component. Great video!
would love to see that more complicated wave motion 🕵
Bro these videos are amazing where have you been in my yt feed all this time
Now you need to turn it to the wine.
Are you planning on covering refractions and depth effect in a future video? It would be really nice followup
in an ideal world i will inevitably cover every topic ever but whether that's within your preferred range of time is another thing
@@Acerola_t lol, yeah that's fair. Depth based stuff is pretty easy, but refraction would definitely take a while.
This is just what I needed for my RTS, only don't need the render, just the waves. It's an aether map, some gameplay stuff, so I'll need to figure out how to get back the data to the cpu eventually. Thank you very mucho!!
Screen space reflections still beat cube maps but are less consistent, I like valves idea of using both screen space reflections and cube maps, mixing methods. If an improvement leads to predictable edge cases, plug the gap with old faithful.
It is always nice when a new Acreola Video rolls out
New favourite acerola video. One day I will be epic like you!
I've been learning opengl recently and i wrote my own code for a wave, but it was just one sine wave, and the vertices were...cubes. And i was computing the position in the cpu. Never thought i should do it in the gpu, that's brilliant. Thanks aceroler
"turn that poop, into wine" - L. Ron Hoyabembe
0:59 but there are real time fluid simulations look at liquigen. It runs on the gpu i think and work in realtime with fairly good performance depending on the scale of the scene.
when i say stuff like this I mean in the context of games.
you have to have the rest of the game too, you cannot dedicate 105% of your resources to a single small asset.
@@Acerola_t yea that's true but I can definitely see it or something similar coming to games in a couple years. Like something similar to portal, not with a whole sea but for puzzles where you use water physics to solve them. That would definitely be very cool and probably not that computationally intensive to run.
The Junes music genuinely makes sponsors fun to listen to
I don't understand the fancy letters and maths but it's close enough to a monogatari episode for me to watch
So I’ve been watching you for a while and I’ve always loved watching these videos despite me barely understanding anything. This was the first time I watched an acerola vid and felt like I actually understood most of it, so much so that it motivated me to download unity and learn shader coding to replicate this water effect. I actually managed to get a nice looking sum of sines with lambertian diffuse and blinn-phong specular! I’ve never touched graphics programming or even a game engine before and somehow this video single handedly dragged me past the barrier. But I’ve kinda been learning exclusively shader programming and avoiding learning Unity itself, apart from like, “Shader Lab” *shudders* Now it seems I actually have to learn Unity to make a script that automatically generate waves with decreasing amplitudes and wavelengths instead of entering those parameters manually
This makes me super happy to hear! If you didn't you should definitely try reading the references in the description to sorta absorb the theory through your applied experience, that's what works best for me.
There's also plenty of ways to extend the project by either doing what you said and trying to create a sort of automated weather system that can blend between different sums, or trying to layer multiple sums of differing scales to reduce the tiling dilemma, or experimenting with different waveforms, or improving the water shader itself by implementing a more sophisticated physically based shader or maybe trying to optimally stylize instead, or perhaps you wanna try and simulate buoyancy and float some objects in the water (it's pretty easy!). It's a super exciting space to work in!!!
Acerola: [explaining concepts beyond my pea-brained comprehension]
me: yooooo he's saying those words Sho Minamimoto says all the time!
Holy moly it’s another acerola video it truly is a wonderful day