Hello, hope you enjoy the video! It was a lot of work getting this far, but definitely still have a long way to go. Clouds are tough! I'd love to hear your thoughts about how they could be improved. Links to all the resources I learnt from are in the description if you'd like to know more. Edit: I made a mistake at 2:12 in saying that the closest point is guaranteed to be inside the adjacent cells, it’s possible to get arrangements where the nearest point is two cells away orthogonally. This doesn’t seem to occur much as I never noticed any discontinuities in the result, but worth knowing. Thanks to everyone who pointed it out.
That extremely well put together and entertaining - and the final result is amazing. Any ideas on how you are going to optimize it? Since you are already doing basically everything on the GPU, what other optimizations exist? Thanks!
It looks awesome, I would like to know if it would be feasible to actually simulate the clouds as they are in real life (as in simulate condensation, evaporation, etc.). It would probably be too much for a GPU to handle so my suggestion would be to hand it off to a server that would then stream the simulation and it's just a matter of rendering.
That endshot looks absolutely beautiful. I'm a programmer myself and hope to be this good in the future. Keep it up because im watching every single video you post
4 ปีที่แล้ว +321
In 2011-2012 I and my friend watched your tutorial about how to make a platform game for our final graduation exam. After 8 years both of us now are technical leader/head of dev but your video still bring new things for us, always! Thank Sebastian, you always a part of our career path.
Scary as hell when going through zero visibility near terrain. Avoid whenever possible. Had to go through some thick fog up in Indonesia once and basically had to trust the console with my life
It better be, the rendering technique he's using is pretty close to physically accurate for how light behaves when passing through a cloud. (He could upgrade his raytracing to photon mapping or even path tracing, which would make it as close as possible, but his GPU is already visibly struggling.)
@@motsgar probably. I dont know much about Blender, but hes got most of the theory in the video so if you follow along and look up the right equations you should get a similar output. The thing I'm not sure about is the shader runs using the compute-shader raytracer he wrote a few videos previously and then upgraded in a subsequent video.
Excellent work! I have 2 suggestions: 1) If it's possible, I recommend adding a 4th dimension into your noise. That way you can do a 4th dimensional offset, which will simulate condensation and evaporation. 2) See if you can send the terrain data back to the noise function. You can then use the heightmap to prevent the clouds from clipping, and maybe even shape the clouds over the mountain like in your IRL video.
You managed to turn something I thought was pretty hard to implement and even understand the papers related is quite hard, into something really easy to follow and achieve with some good patience and tweaking, I also vote for a part 2!
I believe in you guys, we can do this :D lets stop watching youtube and start coding (wich is basicly watching youtube browsing stackoverflow for code help, ow and apparently reading papers)
He is impressive as hell, but most of the stuff he does takes a lot of planning and a shit ton of math. He is definitely confident in his math abilities, but nothing he was programming was too outlandish.
@@xjonnyd93x Agreed. The programming itself isn't so bad. Its really the math and knowing how to approach the problem, which is gained by years of experience.
Well, keep in mind the fact that he dedicates a LOT of time into these. No matter your coding skill, if you put effort amd time into it you can achieve your goal.
You claim it's far from finished, yet looking at the end of your video, my jaw dropped at how real this feels. Sure, there are artefacts and whatnot, but it looks and FEELS like a natural cloud! You, sir, are living proof that if you arm yourself with the right knowledge, you can create living, breathing worlds out of thin air (and code). I'm instantly liking and subscribing!
Except this is a huge plane that is placed on the scene, and won't generate realistic, indivual clouds. Can't create other weather phenominons. Good for demo, but does it work for actual games? It does give me an ides though. Using the same method, but individual 3d planes/cubes, but the plotting at and offset from the edge so the fluff doesn't cut off. Then with these individual clouds, different types of clouds can be places near them, and blend them. ... all very heavy effort. Easy to say, hard to do.
There was a short clip of the real clouds in the mountain in this video. The clouds were behaving very differently (they had physics) while this is just a 3D texture and not simulated with molecular precision (🤓) (or just acting like real clouds and have physics). There is new techniques (🤓🤓🤓🤓🤓) that have better performance and have better clouds. I hate myself.
This is great, i especially appreciate the montage of all the semi failed attempts along the way. For me, this is the most fun part of doing creative code work, searching the space of a new problem and debugging all the strange outputs. It helps this feel less like a rigid tutorial and more like exploring an interesting idea with a friend.
@@gblawrence034 The question is are you ok? All I see on the internet nowadays are stupid kids using this "meme" format in places where it really doesn't make sense just to act cool. Ever heard of using normal sentences?
Holy! Man, this is crazy. I was looking for a tutorial on how to make clouds revolve around a planet (view from a spacecraft in space) just to dolly up one of my low poly planet models. This is way beyond what I need but I'm in awe of the depth and knowledge you have on this. Fantastic work.
Imagine: A fully simulated planet with more detailed, naturally eroded terrain that renders as you fly in, surrounded by birds that act like actual flocks, realistic clouds and functioning ecosystems filled with animals that evolve over time. That would be the single best simulator Earth
"With a little bit of C# coding using a generation function, I've created New York in the year 2021, a pretty relative look back to what it actually looked like using only coding."
I am Brazilian and even without understanding a word of what you say, I feel wonderful and inspired again with my projects, more than an amazing class, there is a great motivator, I confess that I felt a huge desire to give you a hug and thank you that is why. I feel amazed when people have the gift of encircling with such immercivity. Thanks
lol, yeah me: writes a script where all it does is spawn a cube when you click the mouse, which has literally more errors than there are lines of code also me: "let's look up how to code clouds!"
I love your coding adventures! I mean, I understand like, a fifth of what you say, because man am I bad at coding, but it's extremely enjoyable even for someone like me to see the project slowly taking shape :) Great work pal! Have a nice day :)
Hey!! There is a website called Kattis that you can use to improve your programming skills, I definitely recommend it. Also will help with lateral thinking.
To be fair, I dont think this has anything to do with "being bad at coding". Coding/Programming itself is accually quite easy. However, the context or problem and the required math and libraries probably makes this hard to understand and follow without any prior knowledge. Which is to be expected.
Very good result, for performance optimization, you can relay on blue noise technique to decrease the number of samples without compromising visual fidelity. For your next coding adventure, I would suggest you to implement "Atmospheric Scattering". Since you already did a terrain generation, as well as cloud generation. Good luck
hans 1. These are all a part of a technical showcase of coding techniques 2. These are complicated because he is using hard mathematics 3. These are difficult because he is using C# 4. These are not the work of a hobbyist 5. These shouldn't discourage you 6. These aren't fun games 7. You can do better
The amount of effort that goes into creating something that to the untrained eye might seem so simple is incredible. And you’ve wasn’t a sub because you were smart enough to actually do it
As someone with a Masters in Meteorological Physics, this was really interesting to me. It doesn't do a great job of modelling real cloud behaviour, especially in terms of motion over time, but that's understandable since cloud physics are incredibly complex and difficult to calculate. But it did make me understand why a lot of simulated clouds might look and act in the way they do (e.g. mostly confined to a single vertical layer, single uniform horizontal velocity, but good simulation of light effects).
@@Foxtrot2F Which is a shame, though understandable due to limitations in users' computational power. In interactions we can usually detect the difference between something that's physically simulated and something that's visually faked, and it has a huge impact on immersion.
Hey. I just found your channel today, trying to find a video about perlin noises. I just started learning about the whole “noise” concept. I couldn’t help but notice how brilliant you and your works are. Even after studying, I just keep on watching your video, getting motivated and craving for more knowledge! Thank you sir, for sharing your talent for free on the internet.
seb lague: made an atmosphere, clouds, ray marcher, boids, terrain generation, ecosystem, procedural planets, path finding... me: cant stop a cube from falling through the ground
Coming back to this video after 4 years, I was laughing at the complicated math equations at around 0:35. Four years later and I'm doing my master thesis on the topic, knowing damn well each of those equations... Just funny how time can change your perspective on certain things
This is just amazing. I am impressed by your talent to observe nature and reproduce it in code. Your are truly the Bob Ross of Computer Science. Much love from Germany and thank you for your outstanding and calming content.
eh... maybe, but you'd probably be better off using something closer to voxel terrain. I believe one of the (PS2-era) Virtua Fighter or DOA games had levels with snow on the ground that would get mashed and deformed by the fighters' footsteps, and it just used a simple heightmap that got knocked down whenever something passed through it.
@@ClokworkGremlin yeah, a dynamic heightmap sounds like the cheapest solution for sure especially that rudamentary implementation. for a bit more razzmatazz stencil type of shit, you could just project whatever is colliding striaght from the bottom up in orthographic with a hard Z distance limit about as far as you want the deform to be deep. map the things that are further away in Z distance with darker shades of grey to make a heightmap (or lighter, depending on which way you want to go), and map them across the whole rgb range (something like "int color=Math.round(vector.z/zmax)*255"), not overwriting any already set values (essentially projected Z -> B/W color mapping to be short about it). you could even use lookup tables (read: more bump/heightmaps) to make it even cheaper, or or go the other way, and pimp it even a bit more, and allow it to deform terrain other than striaght down,by calculating the inverse angle at which the object approaches it (IE: "player.velocity*=-1" for every axis, and using the outputs to decide where to project the object from, and the factors to which each axis on the target plane should be deformed. (read: the intensity factor for moving the vertices) just mix the 2 maps with regular 2d, as in poking in the pixel data arrays with basic maths (multiplication) could even do with addition/subtraction alone, or even hard setting values, but add/sub wouldn't respect the form of prior deformations as much, unless you made some conditions you can completely avoid using multiplication, and the other (setting) wouldn't respect any old data at all, obviously. but that's merely a matter of preference the geometry may get a bit busy at some point if you want it to look good though, but that's why these things almost always had a hard count limit with modulo on the array index for good old last-in-first-out. or a decay time. tackling that may be a bit more complicated, but i'd opt for something like introducing a view distance falloff for geometry simplification, by keeping multiple mesh/vertex datasets at varying degrees of detail/complexity at hand, using precalculated positions, and bounding boxes of the deformations outside of the range as texture position and size, to fake your way out of it even dynamic unsubdividing is pretty easy if you stick to planes, not harder than calculating pixel or tile array position from X,Y values using the width and height maximums in 2d, as planes have no complexity in the Z dimension whatsoever, it's all still 2d under the hood (akin to "index=(ob.x%w)+Math.floor(ob.y/w)" , or "coord_x=Math.round(ob.x/tilesz)*tilesz" and so on...) all i know is, i would avoid dynamic geometry complexity like the plague, changing from anything but a plane will mess up the heightmap mapping unless you use a ton of witchcraft, and just duplicating geometry as a seperate object is entering a minefield of z-fighting and collision detection problems, easily the 2nd hardest approach (and not even all that cpu friendly either if the whole process is to be done in real-time) they may be more balls to the wall once said and done, but getting there... ugh, i'd rather die of natural causes
@@klontjespap you're getting deep into risk of over simulation there, honestly. Depending on the snow depth you're looking to simulate, what kind of deformations, and the level of dynamics you're looking at, one of the linked particle simulations 2-Minute Papers has featured might work.
Absolutely brilliant! I've been wanting to do my own ray marching cloud shader for many months now ever since I was blown away by Red Dead Redemption 2's clouds. This is a great guide to get me started :D. I'm very curious to see how you tackle the performance limitations. Love the videos Sebastian, your style is phenomenal, keep it up. Nice to see a fellow South African producing such quality content.
Seeing how a slider were used to move the clouds, I started wondering about maybe doing the cloud-generating algorithm not in 2d as initially explained, not in 3d as it moved to later, but in 4d. So points are randomly distributed as before, but this time instead of vec2 or vec, it would be vec4(x,y,z,w), and then for each possible point (still vec4) in structure the distance to closest point is encoded like before. But when rendering, only a 3d box is rendered as before (lock the w-value), and as time passes the box is replaced by "moving" through the fourth dimension. This should make the clouds morph in a continuous manner, never going discontinuous, although might look a bit odd for clouds as they might "morph" into existence where it looked to be relatively empty, or out of existence where it was dense (as if water condenses and evaporates out of thin air).
@@SirMonkeybutt10 I thought of it myself, and I realized later that its actually a bad idea for clouds (technically, I realized it halfway into writing it, as the last note shows, but by then I had already invested effort in writing it). Granted, the idea is sound if you want a continuous morphing geometry, but that is decidedly _not_ how clouds move irl :p I also have no idea how it would actually look, as you would either have 4-dimensional triangles that would somehow need to be rendered, or be treating the fourth component as solely a time-axis (and in that case need to re-generate the geometry each step in time).
@@TestSubject06 this is a bit of a necro, so my memory is probably wrong. But was this video not about using cubemarching over 3d noise to generate geometry enclosing volumes where the noise passed a certain threshold? And then using this geometry/mesh to render "clouds"?
This is actually how many graphics softwares evolve noise so that it smoothly changes to a different look, though I'm impressed you came up with it on your own.
As a programmer myself, I found this to be highly educational. I'm not a game dev, solely work on Mobile Applications and Software, I now understand why TF graphics cards have to work so insanely hard. That's impressive and I've often considered making something of the sort myself, but with rain and collision on the particles. Kudo's to you earned a sub.
You make it look so simple, and yet, to understand what should be the correct mathematical algorithms and how to implement them... Requires a genius mind. Absolutely stunning, well done!
next video: Coding Adventure: Universe Hey, last month I was studying how the universe formed together, so I decided to make one on my quantum computer, check it out
Beautiful! With this rich set of parameters, a game could really implement a fluid weather-system. Just imagine, standing at some exposed are, when suddenly the sky starts turning black.
One of your coolest videos. Another interesting graphics idea are Shadows. You can get really deep with that stuff. Shadow techniques/algos to create nice realistic soft shadows, others to capture contact shadows (small shadows not picked up by normal shadowmap techniques) and so on. It would be really interesting and unique look especially with your amazing style of videos. Also... would be cool if you do this stuff in URP, since the built-in/legacy render pipeline is in fix-mode only, the main render pipelines are URP/HDRP now. URP considered the replacement for built-in and main render pipeline now.
I discovered your channel today and I'm amazed. You are professor material, the way you explain and visualize things for us makes it pretty easy to follow. Your editing also helps a lot. Keep it going!
Go to Google and type in 'C# docs tutorial'. Same with TH-cam, and you'll find a series of tutorials showing you how to programme in C#, and when you get a base underatanding (this takes time) you can play with Unity for free, which also has free tutorials on how to C# code with Unity.
Sebastian you are a blessing! i'm currently in a game dev education and a lot of my fellow programmers follow your videos for inspirations and guidance for how to tackle and implement such features!
I love these videos so so so much. I was never interested in game development at all until I started watching them. Now I'm very strongly considering downloading Unity and trying it out. I know it's a lot of work, and that these videos are just the highlights, but man, this is inspiring!
Just discovered this channel from this video. I gotta say - the topics covered, explanations and presentation are bloody amazing. I'll be sticking around
This looks great! However, with clouds they actually also obscure light. Maybe in a future video you could try to make these clouds cast shadows? Not important, but would be an interesting challenge for sure.
This is absolutely amazing. I especially appreciate that you post the resources that you have found. Please keep doing that, as they seem very helpful, and I may use them in the future myself.
I've always wondered if it was possible to procedurally generate clouds but never bothered looking it up expecting it to be extremely cryptic and difficult to understand. I never expected my question to be answered in the form of a neatly packaged and entertaining TH-cam video. Great job :)
"Messed around till I got what I wanted as a result " said every good coder ever..... bravo the clouds at the end were amazing. Coding sure has changed since the PDP-1170 with the Hazeltine 1500 monitor. ... I will leave that there to tell my age.... lol
Duuuudeeeee so simple and so far so much things you can do with this simulator! Really love the series, the patience, the voice, and the project!!! Looking forward to see how it goes! Keep up this good work! Shared again of course 😊
11:50 Idk why the scene looks like something out of the mountains of madness if it was a movie. Know what, I’m gonna take it all and use it to design my own clouds, and I will have terrain generated by a hydraulic system, I will have animals that will live in an advanced ecosystem... but they can also fall off the world because I just want to see things falling... and I will use boids for herds and flocks of animals.
Each time you upload a new video, I first read the title and find myself thinking "I can tell what he's going to do, I've dabbled with this stuff before". Then I watch the video and realize that your work is so much more elaborate than everything I ever did. Great job as always!
Every time I come across one of your videos and watch it front to back, I think to myself "I should subscribe to this guy, he makes wonders" and then I remember. I already did.
I really like these Coding Adventures. It's interesting and fun. Unfortunately I'm not a C# programmer so I don't understand some parts. (I'm a C programmer.)
@@elektra81516 If you're a Java programmer, yes, it's easy to switch to it. It has better performance than Java due to using a compiler rather than an interpreter. I don't know where I could C#, so I don't really know much. Though, I haven't seen any C# compilers for Linux either.
The end result is beautiful, and your editor setup is brilliant. Having goods tools like these is extremely satisfying :) I'm very late to the party but I can't sleep so... some random thoughts about how this could, maaaaaaybe, be optimised: If your light source is static and directional, your second ray marching (7:45 from inside the cloud to the sun) could be precomputed and baked into one channel, or computed in parallel over several frames if we have a slow moving light source (but still directional). Iirc, a version of sphere tracing can also be used here to speed up the ray marching, if the noise function is continuous (and probably some other prerequisite(s) that eludes me right now...), you can use it to compute a safe bound for your leaps, instead of a fixed step. Then, once again if the values we baked earlier have some nice properties, approximate the transmittance by using samples of this 3D texture at both end of the step and the step length. And to reduce the number of sample needed, you could try gathering the results from the neighbouring pixels after offsetting them by a fraction of a step. I tried this while experimenting with volumetric lights and it really helped my quality/performance ratio... even though it never reached a usable state :D I have no idea how one would go about adding some form of collision with the terrain without it being extremely taxing (except by using the heightmap to fade the coulds away, but that's boring!), but I would really enjoy seeing these clouds roll over the mountains :D
"Obviously, there's still so much more to do! " **By the time, he created a superior version of Microsoft's flight sim, from scratch, in one week or so**
The flight sim he made is not: -superior to Microsoft's: it has a very limited area, only one plane which only has one function, very monotonous terrain, etc. -from scratch (he said he took a premade plane model and controller from the internet and he used his own premade terrain generation) It does have amazing visuals though, on which an asthetics-based (rather than function-based) sim could work on.
@@gnupfo I meant it as a joke XD Although I think the clouds actually surpass the ones of micr. flight sim (Yes, I do know that the graphic power needed is too much to integrate into a flight sim) have nice day or should I say, cheers
Performance idea: when calculating the brightness for a pixel, you might be able to cast every other ray normally, but when you cast the remaining rays you might interpolate the light data from the surrounding properly calculated rays to approximate the light data for the current ray.
I don’t know why it took me so long to realize you were trying to make your clouds as realistic as possible. I think my mind automatically goes to abstract or cartoony when I think about either clouds or water in games. This is awesome!! A long way away from the Minecraft clouds I’m used to, hahaha
You are great, man! i did game dev (self-study) for a while but i stopped because i was overwhelmed by all the things involved (scripting, sounds, graphics etc. ) , yet it's really inspiring to see u build games from scratch.
This is procedural code made buddy...that means you can make, control and change/chain it with any of the infinite possibles and shapes, colors n sizes the world has to offer in real time....that means you can do sky/cloud timelapses for video , weather sim and games ofcourse because you have probably made a clouds sim program with this.
The universe is just quantum mechanics applied to a large scale. Quantum mechanics are probabilistic, which means when two systems are in the same quantum state, observing them may yeald different results. Perlin noise however is computed using a known algorithm which allways yealds the same result if you input the same parameters, therefore it is not truly probabilistic but deterministic. So the universe can't be made of perlin noise.
Hello, hope you enjoy the video! It was a lot of work getting this far, but definitely still have a long way to go. Clouds are tough! I'd love to hear your thoughts about how they could be improved.
Links to all the resources I learnt from are in the description if you'd like to know more.
Edit: I made a mistake at 2:12 in saying that the closest point is guaranteed to be inside the adjacent cells, it’s possible to get arrangements where the nearest point is two cells away orthogonally. This doesn’t seem to occur much as I never noticed any discontinuities in the result, but worth knowing. Thanks to everyone who pointed it out.
Love your videos!
Cool videos!
That extremely well put together and entertaining - and the final result is amazing. Any ideas on how you are going to optimize it? Since you are already doing basically everything on the GPU, what other optimizations exist? Thanks!
That looks absolutely stunning! You inspire me so much😄
It looks awesome, I would like to know if it would be feasible to actually simulate the clouds as they are in real life (as in simulate condensation, evaporation, etc.). It would probably be too much for a GPU to handle so my suggestion would be to hand it off to a server that would then stream the simulation and it's just a matter of rendering.
It was quite a shock seeing my aircraft controller randomly show up in this video! I'm still laughing. Thank you for the shoutout!
subbed to you awesome content on your channel.
Hello fellow DCS player
@@level5822 Thanks a lot for mentioning that game name I've finally found the kind of game I've been looking for since months
^-^
@@Beengus I know english. you know not
@@lukelader It was a joke you fool. Couldn't you tell, you tool?
"Every day the clouds would float past my window, taunting me."
Hahaha, love your videos!
Programmer yells at cloud
I thought this was an inspirational quote
Lol
We all saw the video. Dingus
@@radthadd no need to be mean
That endshot looks absolutely beautiful. I'm a programmer myself and hope to be this good in the future. Keep it up because im watching every single video you post
In 2011-2012 I and my friend watched your tutorial about how to make a platform game for our final graduation exam. After 8 years both of us now are technical leader/head of dev but your video still bring new things for us, always! Thank Sebastian, you always a part of our career path.
As a pilot, can confirm that the end is pretty damn accurate to what flying in a cloud looks like.
Scary as hell when going through zero visibility near terrain. Avoid whenever possible. Had to go through some thick fog up in Indonesia once and basically had to trust the console with my life
@@leaderofcommunistchina1427 man, you have some hardened steel balls
It better be, the rendering technique he's using is pretty close to physically accurate for how light behaves when passing through a cloud.
(He could upgrade his raytracing to photon mapping or even path tracing, which would make it as close as possible, but his GPU is already visibly struggling.)
@@ClokworkGremlinwould it be possible to write a shader like this in blender and make it look really good?
@@motsgar probably. I dont know much about Blender, but hes got most of the theory in the video so if you follow along and look up the right equations you should get a similar output.
The thing I'm not sure about is the shader runs using the compute-shader raytracer he wrote a few videos previously and then upgraded in a subsequent video.
The ending is absolutely gorgeous.
The little touches like the plane leaving wing trails in the clouds are really cool.
I actually didn't notice that
@@kurtisgibson2929 me either.
12:18 and 12:22 are where it's visible. I don't know if that came with the plane/controller simulation, or if Sebastian did that himself.
Excellent work! I have 2 suggestions:
1) If it's possible, I recommend adding a 4th dimension into your noise. That way you can do a 4th dimensional offset, which will simulate condensation and evaporation.
2) See if you can send the terrain data back to the noise function. You can then use the heightmap to prevent the clouds from clipping, and maybe even shape the clouds over the mountain like in your IRL video.
MrMusAddict Interesting ideas, thanks!
i would love to see this implemented. love the video @sebastian
can't wait for Coding Adventure: Clouds 2
@@SebastianLague Would love to see a follow-up in later videos!
You managed to turn something I thought was pretty hard to implement and even understand the papers related is quite hard, into something really easy to follow and achieve with some good patience and tweaking, I also vote for a part 2!
No matter how good I think I’ve become, I watch something like this and feel like I’m coding in crayon.
same man this guy make feel waaay too stupid
I believe in you guys, we can do this :D lets stop watching youtube and start coding (wich is basicly watching youtube browsing stackoverflow for code help, ow and apparently reading papers)
He is impressive as hell, but most of the stuff he does takes a lot of planning and a shit ton of math. He is definitely confident in his math abilities, but nothing he was programming was too outlandish.
@@xjonnyd93x Agreed. The programming itself isn't so bad. Its really the math and knowing how to approach the problem, which is gained by years of experience.
Well, keep in mind the fact that he dedicates a LOT of time into these. No matter your coding skill, if you put effort amd time into it you can achieve your goal.
You claim it's far from finished, yet looking at the end of your video, my jaw dropped at how real this feels. Sure, there are artefacts and whatnot, but it looks and FEELS like a natural cloud! You, sir, are living proof that if you arm yourself with the right knowledge, you can create living, breathing worlds out of thin air (and code). I'm instantly liking and subscribing!
Samewise Gamgee. Super duper impressed.
Except this is a huge plane that is placed on the scene, and won't generate realistic, indivual clouds. Can't create other weather phenominons. Good for demo, but does it work for actual games?
It does give me an ides though. Using the same method, but individual 3d planes/cubes, but the plotting at and offset from the edge so the fluff doesn't cut off. Then with these individual clouds, different types of clouds can be places near them, and blend them. ... all very heavy effort. Easy to say, hard to do.
There was a short clip of the real clouds in the mountain in this video. The clouds were behaving very differently (they had physics) while this is just a 3D texture and not simulated with molecular precision (🤓) (or just acting like real clouds and have physics). There is new techniques (🤓🤓🤓🤓🤓) that have better performance and have better clouds.
I hate myself.
I love watching this guy be good at coding instead of using my time to actually learn coding.
Same
This is great, i especially appreciate the montage of all the semi failed attempts along the way. For me, this is the most fun part of doing creative code work, searching the space of a new problem and debugging all the strange outputs. It helps this feel less like a rigid tutorial and more like exploring an interesting idea with a friend.
I totally agree with you!
and sometimes one of your fails is good enough to maybe be a feature in something else (;
Boids + Clouds + Weathered terrain + simulated fauna + planet generation
... equals an unhappy cpu
Came here to say this! He should definitely try this!
once terrain generation is done the weathered terrain does not affect performance much...
Imagine when he makes the clouds Interact with the mountains...
Lol GPU*
The interstellar survival game he has the tools to make will be amazing.
Sebastian: makes clouds
me: woah
Also Sebastian: casually makes it into a flight simulator
me: wtf
Is this supposed to be funny, relatable, quirky or whatever you normies think is cool nowadays.
Aqua Armour are you okay?
@@aquaarmour4924: when my face when tfw'd
me: pikachu pogger emoji
@@gblawrence034 The question is are you ok? All I see on the internet nowadays are stupid kids using this "meme" format in places where it really doesn't make sense just to act cool. Ever heard of using normal sentences?
@@aquaarmour4924 you seem extremely angry at literally nothing
Holy! Man, this is crazy. I was looking for a tutorial on how to make clouds revolve around a planet (view from a spacecraft in space) just to dolly up one of my low poly planet models. This is way beyond what I need but I'm in awe of the depth and knowledge you have on this. Fantastic work.
for some reason I am both crying & amazed. this was beautiful
Sorry but I can't like ur comment
bcs..
it will ruin the number
Ok great
i know why you are crying
420 likes lol
@@rfisty what?
Imagine:
A fully simulated planet with more detailed, naturally eroded terrain that renders as you fly in, surrounded by birds that act like actual flocks, realistic clouds and functioning ecosystems filled with animals that evolve over time.
That would be the single best simulator Earth
microsoft flight simulator
Would destroy any computer we could have 😂😂😂
look at microsofts flight simulator, it renders the entire Earth without loading screens.
@@Danuxsy yh, but it doesn't have animals or ecosystems does it?
That was the main point I was getting at
@@bobthedeleter no but it does have all the other things you mentioned.
Sebastian Lague in 2050:
"Coding Adventure: The entire Universe"
Quantum Physics
~~I'd pay to watch that~~
I'd pay to install that
"With a little bit of C# coding using a generation function, I've created New York in the year 2021, a pretty relative look back to what it actually looked like using only coding."
Depends on what you are emulating but it's entirely possible
"Coding Adventure: Life" xD
I am Brazilian and even without understanding a word of what you say, I feel wonderful and inspired again with my projects, more than an amazing class, there is a great motivator, I confess that I felt a huge desire to give you a hug and thank you that is why. I feel amazed when people have the gift of encircling with such immercivity. Thanks
BEAUTIFUL gameplay at 11:45, props to you. The atmosphere is incredible, wow.
I've never had the chance to experience such tranquillity in a game.
Me: " yeah that works"
*Looks at hello world program with 473 errors*
Me: "well guess i'll watch another coding adventure"
lol, yeah me: writes a script where all it does is spawn a cube when you click the mouse, which has literally more errors than there are lines of code
also me: "let's look up how to code clouds!"
I started with BASIC, print loop: "Hello world!"
Now Im not allowed to use GOTO anymore :( I GOTO clouds and dream. Syntax error
Lol
hahahaha i deinstalled cisual studio because i got so much errors with hello world xD (and on vs u get a sample of hello world code xD)
This is too real!
I love your coding adventures!
I mean, I understand like, a fifth of what you say, because man am I bad at coding, but it's extremely enjoyable even for someone like me to see the project slowly taking shape :)
Great work pal!
Have a nice day :)
Hey!! There is a website called Kattis that you can use to improve your programming skills, I definitely recommend it. Also will help with lateral thinking.
@@committedcoder3352 ill try it
To be fair, I dont think this has anything to do with "being bad at coding". Coding/Programming itself is accually quite easy. However, the context or problem and the required math and libraries probably makes this hard to understand and follow without any prior knowledge. Which is to be expected.
Very good result, for performance optimization, you can relay on blue noise technique to decrease the number of samples without compromising visual fidelity.
For your next coding adventure, I would suggest you to implement "Atmospheric Scattering". Since you already did a terrain generation, as well as cloud generation.
Good luck
No wonder I can't find a job. Everyone else is smarter than me.
- ᔕጠጠ ᓬェᗆ⊣ ⤙o⊂ ᗜ-ᗜ ェጠᓓጠ
Hint: look sideway.
he's been doing this stuff for over 10 years
Do you have a portfolio? If no then there's no chance
Just get zhe Flammenwerfer.
hans
1. These are all a part of a technical showcase of coding techniques
2. These are complicated because he is using hard mathematics
3. These are difficult because he is using C#
4. These are not the work of a hobbyist
5. These shouldn't discourage you
6. These aren't fun games
7. You can do better
The amount of effort that goes into creating something that to the untrained eye might seem so simple is incredible. And you’ve wasn’t a sub because you were smart enough to actually do it
As someone with a Masters in Meteorological Physics, this was really interesting to me. It doesn't do a great job of modelling real cloud behaviour, especially in terms of motion over time, but that's understandable since cloud physics are incredibly complex and difficult to calculate. But it did make me understand why a lot of simulated clouds might look and act in the way they do (e.g. mostly confined to a single vertical layer, single uniform horizontal velocity, but good simulation of light effects).
Oh man. You will have a blast with the new flight simulator Microsoft is making. Check out the videos
I think he deliberately made it crude due to real clouds taking a lot of computational power
Point in gamedev is to fake cloud movement and not to simulate them.
@@Foxtrot2F Which is a shame, though understandable due to limitations in users' computational power. In interactions we can usually detect the difference between something that's physically simulated and something that's visually faked, and it has a huge impact on immersion.
@@Dystisis Not if you're clever about it.
Each time i see your coding adventures i just find it amazing... that looks awesome to fly in this wonderful clouds ! Congrats ! 😁
"We've just created the worlds worst edge detection effect"
I laughed so hard :D
At least I finally got to know how to make an edge detection effect.
Holy crap, those are some of the most legit clouds I've seen in a game engine. I love this. Idk where I would even start with something like this.
Hey. I just found your channel today, trying to find a video about perlin noises. I just started learning about the whole “noise” concept. I couldn’t help but notice how brilliant you and your works are. Even after studying, I just keep on watching your video, getting motivated and craving for more knowledge! Thank you sir, for sharing your talent for free on the internet.
I applaud all of the work you put into these videos and the end result always amazes me. Keep it up, please!
That looks AMAZING
That final shot of you drifting through the clouds literally was so peaceful I think it lowered my heart rate.
That is honestly one of the most amazing projects I've seen, especially rendered in real time! Love your videos!
I love that you always include a quick first person tour of what you create at the end! Beautiful!
seb lague: made an atmosphere, clouds, ray marcher, boids, terrain generation, ecosystem, procedural planets, path finding...
me: cant stop a cube from falling through the ground
Holy... this looks complicated, good work!
You're making me want to start doing coding projects again! I love your stuff its always really inspirational, and interesting!
"..., but every day, the coulds would flow past my window, taunting me."
There's not many places you can use that woah
Coming back to this video after 4 years, I was laughing at the complicated math equations at around 0:35. Four years later and I'm doing my master thesis on the topic, knowing damn well each of those equations... Just funny how time can change your perspective on certain things
This is just amazing. I am impressed by your talent to observe nature and reproduce it in code. Your are truly the Bob Ross of Computer Science. Much love from Germany and thank you for your outstanding and calming content.
Can you do a coding adventure simulating snow on the ground, packable when something moved over it, using something like this?
eh... maybe, but you'd probably be better off using something closer to voxel terrain. I believe one of the (PS2-era) Virtua Fighter or DOA games had levels with snow on the ground that would get mashed and deformed by the fighters' footsteps, and it just used a simple heightmap that got knocked down whenever something passed through it.
@@ClokworkGremlin yeah, a dynamic heightmap sounds like the cheapest solution for sure
especially that rudamentary implementation.
for a bit more razzmatazz stencil type of shit,
you could just project whatever is colliding striaght from the bottom up in orthographic with a hard Z distance limit about as far as you want the deform to be deep.
map the things that are further away in Z distance with darker shades of grey to make a heightmap (or lighter, depending on which way you want to go), and map them across the whole rgb range (something like "int color=Math.round(vector.z/zmax)*255"), not overwriting any already set values (essentially projected Z -> B/W color mapping to be short about it).
you could even use lookup tables (read: more bump/heightmaps) to make it even cheaper,
or or go the other way, and pimp it even a bit more, and allow it to deform terrain other than striaght down,by calculating the inverse angle at which the object approaches it (IE: "player.velocity*=-1" for every axis, and using the outputs to decide where to project the object from, and the factors to which each axis on the target plane should be deformed. (read: the intensity factor for moving the vertices)
just mix the 2 maps with regular 2d, as in poking in the pixel data arrays with basic maths (multiplication)
could even do with addition/subtraction alone, or even hard setting values,
but add/sub wouldn't respect the form of prior deformations as much, unless you made some conditions you can completely avoid using multiplication, and the other (setting) wouldn't respect any old data at all, obviously.
but that's merely a matter of preference
the geometry may get a bit busy at some point if you want it to look good though, but that's why these things almost always had a hard count limit with modulo on the array index for good old last-in-first-out. or a decay time.
tackling that may be a bit more complicated, but i'd opt for something like introducing a view distance falloff for geometry simplification,
by keeping multiple mesh/vertex datasets at varying degrees of detail/complexity at hand,
using precalculated positions, and bounding boxes of the deformations outside of the range as texture position and size, to fake your way out of it
even dynamic unsubdividing is pretty easy if you stick to planes, not harder than calculating pixel or tile array position from X,Y values using the width and height maximums in 2d, as planes have no complexity in the Z dimension whatsoever, it's all still 2d under the hood
(akin to "index=(ob.x%w)+Math.floor(ob.y/w)" , or "coord_x=Math.round(ob.x/tilesz)*tilesz" and so on...)
all i know is, i would avoid dynamic geometry complexity like the plague, changing from anything but a plane will mess up the heightmap mapping unless you use a ton of witchcraft, and just duplicating geometry as a seperate object is entering a minefield of z-fighting and collision detection problems, easily the 2nd hardest approach (and not even all that cpu friendly either if the whole process is to be done in real-time)
they may be more balls to the wall once said and done, but getting there... ugh, i'd rather die of natural causes
@@klontjespap you're getting deep into risk of over simulation there, honestly. Depending on the snow depth you're looking to simulate, what kind of deformations, and the level of dynamics you're looking at, one of the linked particle simulations 2-Minute Papers has featured might work.
Snow on the ground is an extremely difficult simulation. Physically accurate simulations of snow on the ground is an unsolved problem.
Yes ! snow simulation, and also water simulation. This can be great.
Absolutely brilliant! I've been wanting to do my own ray marching cloud shader for many months now ever since I was blown away by Red Dead Redemption 2's clouds. This is a great guide to get me started :D. I'm very curious to see how you tackle the performance limitations.
Love the videos Sebastian, your style is phenomenal, keep it up. Nice to see a fellow South African producing such quality content.
Seeing how a slider were used to move the clouds, I started wondering about maybe doing the cloud-generating algorithm not in 2d as initially explained, not in 3d as it moved to later, but in 4d.
So points are randomly distributed as before, but this time instead of vec2 or vec, it would be vec4(x,y,z,w), and then for each possible point (still vec4) in structure the distance to closest point is encoded like before. But when rendering, only a 3d box is rendered as before (lock the w-value), and as time passes the box is replaced by "moving" through the fourth dimension. This should make the clouds morph in a continuous manner, never going discontinuous, although might look a bit odd for clouds as they might "morph" into existence where it looked to be relatively empty, or out of existence where it was dense (as if water condenses and evaporates out of thin air).
This is actually a really out of the box good idea, did you come up with this or did you see this done somewhere (no offense)
@@SirMonkeybutt10 I thought of it myself, and I realized later that its actually a bad idea for clouds (technically, I realized it halfway into writing it, as the last note shows, but by then I had already invested effort in writing it). Granted, the idea is sound if you want a continuous morphing geometry, but that is decidedly _not_ how clouds move irl :p
I also have no idea how it would actually look, as you would either have 4-dimensional triangles that would somehow need to be rendered, or be treating the fourth component as solely a time-axis (and in that case need to re-generate the geometry each step in time).
@@feha92 The clouds aren't using geometry here though.
@@TestSubject06 this is a bit of a necro, so my memory is probably wrong. But was this video not about using cubemarching over 3d noise to generate geometry enclosing volumes where the noise passed a certain threshold? And then using this geometry/mesh to render "clouds"?
This is actually how many graphics softwares evolve noise so that it smoothly changes to a different look, though I'm impressed you came up with it on your own.
As a programmer myself, I found this to be highly educational. I'm not a game dev, solely work on Mobile Applications and Software, I now understand why TF graphics cards have to work so insanely hard. That's impressive and I've often considered making something of the sort myself, but with rain and collision on the particles. Kudo's to you earned a sub.
You make it look so simple, and yet, to understand what should be the correct mathematical algorithms and how to implement them... Requires a genius mind. Absolutely stunning, well done!
Really nice results especially considering the equally beautiful explanations, thank you :)
next video:
Coding Adventure: Universe
Hey, last month I was studying how the universe formed together, so I decided to make one on my quantum computer, check it out
lol
Beautiful! With this rich set of parameters, a game could really implement a fluid weather-system. Just imagine, standing at some exposed are, when suddenly the sky starts turning black.
Sebastian, I just wanted to say that I love your work. Thank you. Seeing that airplane fly gave me some feels. Your work is inspiring.
One of your coolest videos. Another interesting graphics idea are Shadows. You can get really deep with that stuff. Shadow techniques/algos to create nice realistic soft shadows, others to capture contact shadows (small shadows not picked up by normal shadowmap techniques) and so on. It would be really interesting and unique look especially with your amazing style of videos. Also...
would be cool if you do this stuff in URP, since the built-in/legacy render pipeline is in fix-mode only, the main render pipelines are URP/HDRP now.
URP considered the replacement for built-in and main render pipeline now.
this guy : trying to simulate clouds for weeks.
cyberpunk 2077 : P R E D E F I N E D
I said "YAY!" out loud when I saw that you had uploaded a new video
This is so technically impressive its mind blowing.
I discovered your channel today and I'm amazed. You are professor material, the way you explain and visualize things for us makes it pretty easy to follow. Your editing also helps a lot. Keep it going!
7:00 this entire clip is just Sebastian flexing on us with his stunning backyard view.
You're fucking awesome, the way you breakdown these videos is so perfect. Great work.
Omg i almost understood 1% of what you've done.
lol same.
I know, right?!
Go to Google and type in 'C# docs tutorial'. Same with TH-cam, and you'll find a series of tutorials showing you how to programme in C#, and when you get a base underatanding (this takes time) you can play with Unity for free, which also has free tutorials on how to C# code with Unity.
Lol
@@paulvinova thx!
Use the noise map for the terrain on a map to control the density and height of clouds so you end up with those clouds formations over mountains
Sebastian you are a blessing! i'm currently in a game dev education and a lot of my fellow programmers follow your videos for inspirations and guidance for how to tackle and implement such features!
Can't believe one man is capable of so much detail, great job!
Just seeing the noise diagram he made at the beginning and realizing it was also a voronoi diagram blew my mind
oh YES. now thats something i wanted to see :D
".. I set about implementing this in code" *shows him typing random stuff with his pinky and thumb*
Well how else are you going to keep the right hand free for pressing backspace until you get the characters you need?
Wait, how do you type?
I love these videos so so so much. I was never interested in game development at all until I started watching them. Now I'm very strongly considering downloading Unity and trying it out. I know it's a lot of work, and that these videos are just the highlights, but man, this is inspiring!
Go for it, it's a lot of fun! Happy to hear you've been enjoying the videos :)
Just discovered this channel from this video. I gotta say - the topics covered, explanations and presentation are bloody amazing. I'll be sticking around
This looks great! However, with clouds they actually also obscure light. Maybe in a future video you could try to make these clouds cast shadows?
Not important, but would be an interesting challenge for sure.
You are a genius, love the video
I'd love to see them react to terrain or other colliders. to get a 'clouds rolling over a mountain' effect.
This is absolutely amazing. I especially appreciate that you post the resources that you have found. Please keep doing that, as they seem very helpful, and I may use them in the future myself.
I've always wondered if it was possible to procedurally generate clouds but never bothered looking it up expecting it to be extremely cryptic and difficult to understand. I never expected my question to be answered in the form of a neatly packaged and entertaining TH-cam video. Great job :)
"Messed around till I got what I wanted as a result " said every good coder ever..... bravo the clouds at the end were amazing. Coding sure has changed since the PDP-1170 with the Hazeltine 1500 monitor. ... I will leave that there to tell my age.... lol
I … had my doubts about the realism of this ‘distance from a point’ thing, but the end result look pretty good.
"I am making Flight Simulator 2020 at home all by myself"
Microsoft flight simulator: I don't think so.
This entire process was mesmerizing! 😍 Learned a lot by seeing your approach to problem solving, amazing work. 💪
Duuuudeeeee so simple and so far so much things you can do with this simulator! Really love the series, the patience, the voice, and the project!!! Looking forward to see how it goes! Keep up this good work! Shared again of course 😊
The result was so good, I thought the thumbnail was clickbate!
idk programming so idk why this is in my recommended...
i still watched the entire video..
11:50
Idk why the scene looks like something out of the mountains of madness if it was a movie.
Know what, I’m gonna take it all and use it to design my own clouds, and I will have terrain generated by a hydraulic system, I will have animals that will live in an advanced ecosystem... but they can also fall off the world because I just want to see things falling... and I will use boids for herds and flocks of animals.
Each time you upload a new video, I first read the title and find myself thinking "I can tell what he's going to do, I've dabbled with this stuff before". Then I watch the video and realize that your work is so much more elaborate than everything I ever did. Great job as always!
Every time I come across one of your videos and watch it front to back, I think to myself "I should subscribe to this guy, he makes wonders" and then I remember. I already did.
Are you planning on adding collisions to create effects like 6:55?
I really like these Coding Adventures. It's interesting and fun. Unfortunately I'm not a C# programmer so I don't understand some parts. (I'm a C programmer.)
It doesnt take too long to figure out, unity itself has some great official tutorials on getting started with the language
@@elektra81516 Yeah. The basic logic is similar, except C# has quite a few more functions already made for you.
@@640kiboughttobeenough C# is designed for performance and rapid speed writing, i moved over too it after using Java for a long time
@@elektra81516 If you're a Java programmer, yes, it's easy to switch to it. It has better performance than Java due to using a compiler rather than an interpreter. I don't know where I could C#, so I don't really know much. Though, I haven't seen any C# compilers for Linux either.
Is C object originated? Ive never used it and ive always wondered how it relates. Im a WPF and C# programmer so ive never some across it
One day I'll be good with coding like this guy. "God bless you..."
The end result is beautiful, and your editor setup is brilliant. Having goods tools like these is extremely satisfying :)
I'm very late to the party but I can't sleep so... some random thoughts about how this could, maaaaaaybe, be optimised:
If your light source is static and directional, your second ray marching (7:45 from inside the cloud to the sun) could be precomputed and baked into one channel, or computed in parallel over several frames if we have a slow moving light source (but still directional).
Iirc, a version of sphere tracing can also be used here to speed up the ray marching, if the noise function is continuous (and probably some other prerequisite(s) that eludes me right now...), you can use it to compute a safe bound for your leaps, instead of a fixed step. Then, once again if the values we baked earlier have some nice properties, approximate the transmittance by using samples of this 3D texture at both end of the step and the step length.
And to reduce the number of sample needed, you could try gathering the results from the neighbouring pixels after offsetting them by a fraction of a step. I tried this while experimenting with volumetric lights and it really helped my quality/performance ratio... even though it never reached a usable state :D
I have no idea how one would go about adding some form of collision with the terrain without it being extremely taxing (except by using the heightmap to fade the coulds away, but that's boring!), but I would really enjoy seeing these clouds roll over the mountains :D
Hi! I just wanted to say that I really like what you do! Continue and don't change!
Basically creates MS Flight Sim 2020 in his spare time "Well that's all I have for now" 😐😐😐
"Obviously, there's still so much more to do! "
**By the time, he created a superior version of Microsoft's flight sim, from scratch, in one week or so**
The flight sim he made is not:
-superior to Microsoft's: it has a very limited area, only one plane which only has one function, very monotonous terrain, etc.
-from scratch (he said he took a premade plane model and controller from the internet and he used his own premade terrain generation)
It does have amazing visuals though, on which an asthetics-based (rather than function-based) sim could work on.
@@gnupfo You must be fun at parties.
@@gnupfo I meant it as a joke XD Although I think the clouds actually surpass the ones of micr. flight sim (Yes, I do know that the graphic power needed is too much to integrate into a flight sim) have nice day or should I say, cheers
Performance idea: when calculating the brightness for a pixel, you might be able to cast every other ray normally, but when you cast the remaining rays you might interpolate the light data from the surrounding properly calculated rays to approximate the light data for the current ray.
I don’t know why it took me so long to realize you were trying to make your clouds as realistic as possible. I think my mind automatically goes to abstract or cartoony when I think about either clouds or water in games.
This is awesome!! A long way away from the Minecraft clouds I’m used to, hahaha
You are great, man! i did game dev (self-study) for a while but i stopped because i was overwhelmed by all the things involved (scripting, sounds, graphics etc. ) , yet it's really inspiring to see u build games from scratch.
Instructions unclear: class 5 hurricane crashed my pc
I hate it when that happens
@@josephhacker6508 "I hate it when he does that"
1:37 did anyone else’s brain decide that was a bunch of maggots
Why. I can't unsee it now
thanks you so much i cant unsee it :)
There’s a ton of different types of clouds, though; this one looked amazing, but how would you make, for example, cirrus clouds?
S t r e t c h the clouds
Honestly, I think that's the easiest solution to make other clouds
(You could also play around with transparency)
@Greenpixel ah, that would really help. Good idea!
wow those clouds are amazing, like youve literally made them perfect
Detail Noise Scale and Weight really made the clouds shine! Awesome work, keep it up :)
Yeah sure, just let me...
*Grabs a cloud sprite and slaps it into a project*
Done.
This is procedural code made buddy...that means you can make, control and change/chain it with any of the infinite possibles and shapes, colors n sizes the world has to offer in real time....that means you can do sky/cloud timelapses for video , weather sim and games ofcourse because you have probably made a clouds sim program with this.
@@raw238 yeah I saw the video too, I was just joking about how complex it was
@@s4nt497 yeah agreed...this is not for me atleast not now but the end result is so satisfying if my potato runs it. Anyway Take care
Saw Sebastian's codes
Me: hey i know some of those words meaning
emphasize on "some"
^
Nature is just detailed Perlin noise
Fight me
The universe is just quantum mechanics applied to a large scale. Quantum mechanics are probabilistic, which means when two systems are in the same quantum state, observing them may yeald different results. Perlin noise however is computed using a known algorithm which allways yealds the same result if you input the same parameters, therefore it is not truly probabilistic but deterministic. So the universe can't be made of perlin noise.
My mind is blown completely now. Love your videos, and creativity!
This is one of the most inspirational coding videos I've seen. Great work Sebastian!