Actually when you talk about the glass shader and increase all light path bounces to 256, that's not what I do. You can leave them all at default or whatever values you prefer for rendering. The only bounces I increase are the ones for transparency and I usually get good results with 64 oder 128. You can leave the total bounces number low, because transparency bounces are independent from them. It doesn't even take much longer to render. And more important it also helps with black artifacts in rendering smoke simulations, which in some tutorials you are told are caused by low resolution or bugs when using the adaptive domain. Anyway, so there is no need for using ray length and transparency combined with the glass shader.
Yeah... That's what I said on the video? I borrowed the trick from a blender conference talk that goes a little more in depth. The bounces that affect the problem are generally glossy/transparent, and the overall "trick" have become almost useless now that cycles is faster. The reason I still kept the light path node shinanigans was because alternating between the ray depth and ray lenght masks controlling transparecy, you can improve render times with lots of glass particles (that would be WAY more expansive to render by just increasing transparency bounces) and also you can use the ray lenght to create a mask and remove the reflections of surfaces interacting with difuse shaded colliders (I.e, when mantaflow meshing creates a gap between the water surface and the collider surface, leading to a unrealistic mirror-like surface). Anyway, thanks for the thoughtful correction, I'll make sure to pin this comment so everyone sees it :) The part about increasing everything to 256 was a joke tho. I tought that was pretty clear, but now I see that rambling about 3090s made the whole thing just a dragged out mess and way less comprehensive.
I could never figure out those damn black artifacts in smoke renders. You are my new favorite person. I have looked high and low on every forum and a thousand tutorials for an answer that isn't low res or adaptive domain issues.
Even for smoke it is either alpha, or sometimes weird intersection stuff from when you have several volumes occupying the same space. Shifting the volume objects a little bit so they don't interssect (i.e the grids of voxels aren't exactly aligned) can help then.
Thank you! I realized after finishing the edit that I left a huge chunk of nonstop info without even a breath, but I had no time left to record new versions of some of the explanations. I'll update the description with a project file so you guys can dig around and reverse engineer stuff :)
2:00 Ok so this is how I'd do it: 1. The Brown/Blue Mask: - TextureCoordinate->generated->separte X. Map Range so you get a considerably sharp ramp. - Distort the ramp: Add a MixRBG (Linear Light Mode) after The Texcdoord Node. Connect the second Input with a Noise Texture. Eventally do the same with the Noise itself to get a "Distorted Nosie" along the ramp. 1.1 Add some sploches to the dirt mask with another Noise Texture 2. use this Mask to drive the Density of a volumetric Shader to create the brown dust. 3. For the Surface Material use a Glass Shader but use the Light path Node to make sure it's only Glass for Camera Rays otherwise Transparent BSDF (or use the ray depth trick if it works here) 4. Add Bump - Noise Texture with ca. 0.3 Roughness and high Detail 5. Add some blue Volume Absorption to the Blue part.
Looks like this would work fine without much effort, which is great. When I put that on the video I had a faint idea of using one of those 2d fluid simulators to generate a mask, but it would be optimal to use blender's own procedural textures.
Thank you for this tutorial! It genuinely helped me because alot of the other techniques i see to make "realistic'ish" mantaflow water is very hardware expensive which leads to very long bake times, but this helped me considerably, so thank you very much!
I think i found amazing youtuber..! I’m just staring out, I can’t keep up with your video. But your amazing works motivate me a lot..! Please keep it up🔥🔥❤️❤️
This made me want to continue to learn blender. I like the way you explain. I was Lost at times because I don't know crap...but this inspired me to try to learn blender once more. Thank you.
Hours of banging my head against the keyboard while listening to blender conference talks and trying overambitious projects. the glass/light path thing is almost certainly from one of the talks (I'll link it if I find it) and probably the point density thing too (maybe the nebulae talk by Gleb Alexandrov?) pro tip: the blender conference talks are a great source of blender knowledge.
Nope. it's the meeting of the waters on the amazon river. both rivers are fresh water, just different enough on temperature and flow speed to go unmixed for miles. Fun fact: the amazon river is so large, and so much water flows within it, that when it meets with the ocean, the change of the tides causes a monstrous wave flowing up the river, trashing everything on it's way. It's called "Pororoca" by the locals, and also the source of a huge meme here in Brazil, the "corre richard" meme... which is basically a TV reporter (who graduated with honors at the Prometheus school of running away from things, btw) getting a little too close to the wave and almost dying in the process. Brazil is hardcore even on meme material lmao
(Not so) fun fact: a lot has changed within blender since I begun to plan this video and even after I finished editing it. (namely, the full repercussions of cycles x and a few changes on how a few nodes work and how blender handles meshes/shading). I also got a lot wrong and learned better ways to do basically everything I talked about on the video. this comment section is filled with good tips too! maybe I will make a updated, better version of this, including some simulation stuff that I gathered over the years fighting mantaflow to get it to behave, but since the theory section is still valid, I decided against deleting it. plus it has made me a full fat 10 bucks so far. jokes aside, principled bsdf is in theory more physically accurate, but I (and a bunch of other geriatrics) still find it faster to just mix transparent and glass bsdf ( sometimes taking fresnel into account, but most times not). principled volume, tho, I'm 100% with you on that. way better than five different volume shaders and lots of sliders.
actualy I meant "volume shaders" as it's own thing, I'm not sure how the principled volume works, although maybe it would produce better results, not sure. (also the original comment was made before I got to the part where you were talking about deeper waters so I didn't know at the time you already mentioned volume shaders)
Funny you just said that, I was considering a waterfall as the example sim, but ended up going with weird rock. I won't promise anything (as my record isn't the best when it comes to deliver promised tutorials) but definetly will try. I love waterfalls, and there's lots of things I want to do that would go well with waterfalls too.
Have you considered using a glass shader with an IOR of 1/1.333 (the inverse of water's) for the bubbles? I found that it realistically simulates a pocket of air inside water. Try it! Also, if you can get away with only spherical partilcles, you can use the new point cloud rendering system, it's much faster than rendering instanced icospheres.
Oh, you gonna hear about point cloud rendering, all right. I'm still mad nobody told me sooner. I've been spending my every awake moment testing the limits of point cloud renders for the past month
Also I did think about the glass bubble thing, but discarded it due to render times. My strategy for bubbles now is a metallic shader with the right ior, with the alpha channel being controlled by the fresnel output on the layer weight node, plus some math nodes and sometimes a very slight white emission with low strength. The downside to this approach is the absolutely massive amounts of transparency bounces needed, to the point of being unpractical beyond a certain threshold on the particle count. But the same applies to the glass shader, and this looks better imo.
Principled Volume shader. Ok, There's a great tutorial by polyfjord, called 'how to make fireballs on blender' that's one of the best fire/smoke tutorials out there. You should guve it a look.
Thanks! I currently don't know of any reliable way to get fast caustics to work with cycles. I can think of a few ways to use procedural textures to fake them, but none of them will look great... maybe sometime in the future? idk.
@@FractalParadox Yes, yes. Just bought it brand new a year ago, monster GPU and upgraded to 128GB RAM also. Only to find out Blender doesn't support MAC GPU's. But just read yesterday Blender 3.1 should support my GPU.
Only a selected few builds have the new caistics tho. The 3.1.2 sable build doesn't have it. Dunno why, maybe it's an issue with experimental features?
@@FractalParadox Just little bit buggy. I'm using CPU btw... 🤣🤣🤣🗿🗿🗿 But not laggy, just low render. Because Ii don't have good GPU and it's expensive right now.
No naming names, it's all in good fun :) Actually I didn't mean for this joke to sound so specific, but puget systems was the first thing that came to mind when I recorded ( and also this was my best take, so...)
Sorry, I just gotta ask, how did you get the water to push the objects at 15:55 like that? I've always wanted to do that, but to my understanding, that was impossible to actually do, you could only fake it by either keyframing it or setting the surface after simulating to a rigid body, but there would be a disconnect with the particles and objects. How did you do that, and did you just fake it super well? I gotta know! Also, great work with the tutorial! I've seen every fluid sim tutorial there is in the world, and I think this one might be the most useful and well-done one out there.
Faked it with rigid bodies. I baked the rigid body sim to keyframes, chose a keyframe where the rocks seemed settled (more or less in the middle, right after they touched the ground) then deleted all the previous keyframes. Took a couple of attempts to get the timing of the water just right, which I did by altering the inflow object's starting keyframe and the overall simulation speed
The foam part isnt working. i plug in the point density as shown but nothing happens. even tried to use a math node to multiply by absurd numbers but the white parts just never show up
Maybe somwthing changed from the blender version I was using (~2.8) to now. Remember: the particle system is end of life right now, and they might have just dropped support to it at some point. I'd try it with an older blender version, then if it still doesn't work, just drop it altogether. Geometry nodes points offer such better performance anyway, that you might nor even need such tricks anymore
@@FractalParadox Well that's unfortunate. Thank you for the great guide anyways. Is it possible for you to give me a few tips to maybe create foam using geo nodes?
I talk more about it here: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-Xg4UpGRVsmo.html if you are using Flip fluids addon, it's points automatically I think. if mantaflow, I show how to render as points in the link above
Eevee is a whole different beast. You would have to do a bunch of things manually cause the light path node don't work on eevee. The volumetric stuff would still work, but the glass amd particles would need special materials and a lot of tweaking
@@PAGMAOnline I'm sorry, but I really thought I had a link to the full shader in the description. Like I could swear on it. But it's not there. Whoops. It's been like a whole year. Wtf
Thanks! You might be surprised how much 20min fiddeling with settings on mantaflow makes you even more confused. But after 200 hours, it starts to make sense
I saw in a comment that you explained how you made the rocks move with the water, but I didn't quite understand the explanation, could you elaborate more for me? I also wanted to know the ideal amount of samples for rendering
The rocks move with a simple rigid body simulation that Ibaked to keyframea and than deleted every keyframe before the water "hits" each rock. The ideal amount of samples will always be however much your time budget allows for. Let's say you have 200 frames to render and 12h. You divide 12h by 200, which would be 0,06h or 3,6 minutes per frame. Let's say minimum samples are about 30. The you set it for something average like 256 or 128, set the time limit for about half of the time you calculated (frames take a lot more time to render than whatever you set on three-time linit' section), and there you go.
I didn't. I simulated the sculpture falling, then added the water later, setting the start of the sim just right so that the water hits it just as it starts to move
No. The transparent/water color not being 100% white can make it more noticeable, but the cause of the problem is light getting "trapped" inside the glass ( since computers can only calculate a set number of bounces instead of the infinite bounces irl). The ray gets in, and bounces back and forth between the internal geometry of the glass object, until the maximum number of bounces is reached. If the ray is still inside of the glass when that happens, it is counted as "occluded", and if there are alot of them in the same area, the average will be darker. Try experimenting with reducing bounces on the different types of ray on the light path settings and see what type of results you get. The worst offenders on this case are generally transparency, glossiness and transmission rays, but increasing the bounces too much makes render times awfully long so that's what the trick is about :)
Hi there , thanks a lot for sharing this. Is very difficult to make the head falling as the water moves? I dont know how to make the water interact with objects and make them moving. Thanks.
Not really. The water isn't actually moving it, I'm just using a rigid body sim to make it look like the water is moving stuff. I baked the simulation to keyframes, then erased the keyframes till the frame when the water hits.
Maybe your scale is not applied? It can disappear if thw scale is wrong. Also the point density is very hit and miss on the viewport, so sometimes it doesn't show, but in the render it's there. Also try selecting the node and changing some parameter to refresh the node tree or restarting blender.
Sadly my monitor res is lower than 1080p. I try to make things as big as possible during screen recordings, but sometimes I forget :P I'm planning on getting a new monitor, but prices in my country are... less than desirable right now. so yeah, lowres it is
@@s0up_dev greater than is probably around .500, layer weight is probably default. You can mess with the values till it looks right. Also alpha will cost you, so you can also use the same setup to make the bubbles transition into a diffuse or metallic shader instead of transparent
i was wondering why it didnt look exactly realistic and i see that its because the stone are not getting darker where tthe water hits it... have any idea of how to do soo ? thanks
Too much hassle for me with this tutorial specifically (as it is about the water shader only), but yes, you can do that traditionally using dynamic paint, or I guess you can use geometry nodes with the geometry proximity node and some material switches. Dynamic paint is a modifier you can use to make certain objects act as "brushes" and "paint" on a "canvas" object. If you set the water mesh as the brush and the rocks/ground as the canvas, you can create a "wetmap" (a bw image that tells you where the "paint" is on the "canvas") that you can use as a mask to mix between your normal shader and a darker, glossing version.
Please, do not add text that cannot be read without pausing the video. As a rule of thumb, try to read it 3x while watching at normal speed. If you cannot, it's too much text, too fast.
Seems like a good rule of thumb... If there was anything useful on the text anyway. Most of the quick text is intentional, and some is just complementary info.
@@henryhambardzumyan8102 oh I'm not angry or anything, the video is super confusing and random, plus my humor isn't the best, and can be very annoying. Plus since 3.1 is here, and a bunch of features got added, it's already outdated info. There are a lot of mantaflow tutorials out there which are more accessible, tho I haven't seen many that tackle shading
@@henryhambardzumyan8102 the behavior of some shaders changed a little with cycles x, and it got a lot less forgiving with very thin overlapping geometry too (you can guess what that does with particles) plus now there's point cloud rendering, so the parts about shape of the particles are also outdated coz it takes less resources to render as a point cloud
You always get them with mantaflow sims tho? Try following a mantaflow sim tutorial, there are lots. Ch3ck if perhaps you are not using the " combined export" option, that merges all particles together
@@FractalParadox How would you do this, I am having trouble getting the shader to work in 3.5 and am wondering how/where/what the heck is a point cloud. If you could help that would be great!
@Samuel Styles point clouds are a new way to render virtualized 'particles' with geometry nodes. I explain how they work better in this video: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-Xg4UpGRVsmo.html
