WEBVTT 00:00.000 --> 00:03.799 Blender 3.6 has been released, and it's including a flashy 00:03.800 --> 00:06.280 new feature that will probably get some heads turning. 00:06.980 --> 00:07.980 Simulation nodes. 00:08.500 --> 00:12.540 Yes, simulation has come to geometry nodes, and in this video I will give you 00:12.541 --> 00:14.675 an overview of what to expect from the new 00:14.676 --> 00:17.060 system and how you can get started using it. 00:17.680 --> 00:20.280 There is one thing I want to make clear from the beginning though. 00:20.640 --> 00:23.675 With Blender 3.6, nothing changes in terms 00:23.676 --> 00:26.900 of how physics simulations are done for now. 00:26.901 --> 00:32.180 Rather than taking all of Blender's existing physics simulation systems and 00:32.181 --> 00:35.352 plugging them into nodes, the new feature does 00:35.353 --> 00:38.940 something much simpler, but also much more powerful. 00:39.600 --> 00:43.780 The approach is to build on the vast capabilities that geometry nodes already 00:43.781 --> 00:49.880 have, and add basic functionality to turn all of that into custom simulation systems. 00:49.881 --> 00:56.340 That way, we make sure to build a solid foundation that power users can benefit 00:56.341 --> 01:00.580 from immensely already, and add more high -level features on top of that, 01:01.200 --> 01:07.020 that any artist can use easily later on, in a way that supports and strengthens the 01:07.021 --> 01:09.680 entire ecosystem of geometry nodes in Blender. 01:10.360 --> 01:14.420 So yes, right now I should give the disclaimer that you do need some solid 01:14.421 --> 01:18.280 experience using geometry nodes to benefit from the new feature. 01:18.281 --> 01:21.943 Easier-to-use, one-click solutions using the 01:21.944 --> 01:25.741 new system will follow in future releases. 01:26.100 --> 01:28.610 But hopefully I can give you an idea how this 01:28.611 --> 01:31.400 approach scales up to virtually infinite potential. 01:32.260 --> 01:37.300 That said, the added feature set is, for now, intentionally capped, 01:37.620 --> 01:41.345 relatively compact, and simple to scale it up alongside 01:41.346 --> 01:45.200 other necessary improvements in Blender in future versions. 01:45.201 --> 01:48.850 And there are some obvious current limitations, like 01:48.851 --> 01:52.200 losing the material assignment, that aren't supported yet. 01:52.720 --> 01:57.340 This is just the first major milestone for simulation in geometry nodes. 01:58.040 --> 02:02.365 But let's take a step back and look at what's actually been added in Blender 3. 02:02.366 --> 02:03.680 6 for simulation nodes. 02:04.500 --> 02:07.235 The new elements in the user interface that hint 02:07.236 --> 02:09.780 at the new simulation system are relatively subtle. 02:09.781 --> 02:13.516 In terms of nodes, there's really just a single new entry in 02:13.517 --> 02:17.720 the add menu to make simulation possible, the simulation zone. 02:18.400 --> 02:22.500 When you add it to the node tree, it creates two nodes that are already 02:22.501 --> 02:26.557 connected with a geometry link and this plum-colored panel 02:26.558 --> 02:29.660 that changes automatically when you move them around. 02:30.920 --> 02:34.578 This panel describes the so-called simulation zone and 02:34.579 --> 02:37.300 it indicates which nodes are part of the simulation. 02:37.301 --> 02:42.220 When I insert the simulation nodes in here, nothing changes at first, 02:42.500 --> 02:46.060 because we are just passing through the geometry without doing anything. 02:46.700 --> 02:53.020 Let's just add an extrude node to make a visible change and use a random value node 02:53.021 --> 02:56.040 for a random selection of faces to extrude. 02:56.840 --> 03:01.980 You can see how the simulation zone adapts its shape to include the extrude node. 03:02.740 --> 03:06.680 The random value node, however, is not part of the simulation itself 03:06.681 --> 03:10.780 because it only provides a field that is used by the extrude node. 03:11.620 --> 03:14.014 The actual data that is part of the simulation 03:14.015 --> 03:16.581 is only passed around between these nodes here. 03:17.240 --> 03:19.480 But so far, there is no simulation yet. 03:20.280 --> 03:24.280 But when I go to the next frame, you can see how this extrusion is done 03:24.281 --> 03:27.480 again, and again, and again, the more frames I progress. 03:27.481 --> 03:33.460 The reason for that is that for each frame that I continue, the result of that 03:33.461 --> 03:38.660 previous frame is passed into its input, creating a sort of loop behavior. 03:39.240 --> 03:43.780 Each individual step of the operation is also called a simulation step. 03:44.260 --> 03:48.940 Okay, great, but I want to have a different random selection for each frame. 03:50.320 --> 03:53.599 So let's go to the random value node and plug 03:53.600 --> 03:56.860 in the frame of the scene into the seed input. 03:56.861 --> 04:00.711 And now when I continue frame by frame, you can see how a 04:00.712 --> 04:03.680 different selection is used for the extrusion on each step. 04:04.240 --> 04:09.120 But the result that we had earlier from the setup, using the same seed for every 04:09.121 --> 04:11.660 frame, is still influencing what we see here. 04:12.220 --> 04:15.620 The reason for that is that the simulation just takes whatever it gets from the 04:15.621 --> 04:17.980 previous frame, no matter what you changed. 04:18.360 --> 04:21.373 So to reset the state of the simulation to its 04:21.374 --> 04:24.140 initial state, we need to go to the first frame. 04:24.141 --> 04:26.923 And now if I play back the simulation, you 04:26.924 --> 04:30.421 can see how this works properly as expected. 04:31.260 --> 04:35.220 We get a different selection for the extrusion on every single frame. 04:35.520 --> 04:40.760 You can also see how in the timeline, playing back builds up this purple bar. 04:41.560 --> 04:44.292 And this bar is representing that the frames 04:44.293 --> 04:46.600 of the simulation are written to the cache. 04:46.601 --> 04:51.680 That means I can always go back in the timeline and get any state of the 04:51.681 --> 04:54.360 simulation without having to calculate it again. 04:55.060 --> 04:58.420 But this is only the case as long as the cache is valid. 04:58.840 --> 05:03.300 If I make any change in the node tree that affects the simulation, the color of this 05:03.301 --> 05:07.164 bar gets slightly more faded, indicating that the cache has 05:07.165 --> 05:10.100 become invalid and the simulation needs to be run again. 05:10.101 --> 05:14.380 But of course, if you want to, you can also just change the values in 05:14.381 --> 05:16.840 real time to affect the simulation that way. 05:18.220 --> 05:21.780 A more permanent way of storing the cache is to bake the simulation. 05:22.300 --> 05:27.000 That can be done in the physics properties tab under the new simulation nodes panel. 05:27.680 --> 05:32.680 This will bake the simulation data for the entirety of the scene frame range and 05:32.681 --> 05:35.460 store it on disk in a folder next to the blend file. 05:35.461 --> 05:40.500 You can also see and change that path here in the modifier panel. 05:41.780 --> 05:45.580 Now that the simulation is baked, you can see how the playback is super 05:45.581 --> 05:48.708 smooth, and when I change values, this will not 05:48.709 --> 05:51.440 actually invalidate the cache automatically anymore. 05:52.500 --> 05:55.511 So in this case, if you want to free it again, you 05:55.512 --> 05:58.680 need to do that explicitly by deleting the cache data. 05:59.500 --> 06:03.520 Let's go over another example, to go a bit more into detail, where I show 06:03.521 --> 06:09.621 you how to build a relatively basic particle simulation setup using simulation nodes. 06:09.780 --> 06:14.220 Note, by the way, how what I've shown so far isn't exactly a physics simulation. 06:15.000 --> 06:18.957 The use cases for simulation nodes really go much further 06:18.958 --> 06:22.220 than what we classically have as a simulation and blender. 06:22.221 --> 06:25.458 First off, let's just create a whole bunch of points, 06:25.459 --> 06:28.640 and for that I use the Distribute Points on Faces node. 06:29.580 --> 06:32.348 With that, we can just create a bunch of points 06:32.349 --> 06:35.861 on the input geometry from the base mesh. 06:36.000 --> 06:39.040 At these points, we then feed into the simulation. 06:39.980 --> 06:42.908 Inside of the simulation zone, the main feature of 06:42.909 --> 06:46.000 the simulation is going to just be moving the points. 06:46.160 --> 06:48.480 For that, we'll use the Set Position node. 06:48.481 --> 06:52.680 Let's just give that a go with just some random offset vector. 06:54.240 --> 06:56.310 And yeah, obviously right now this is not very 06:56.311 --> 06:59.120 interesting, but this is going to be the main concept. 06:59.920 --> 07:01.800 Just moving points in a certain direction. 07:02.520 --> 07:05.004 And to keep track of the direction and speed of 07:05.005 --> 07:08.040 that movement, we're going to add a Velocity. 07:08.620 --> 07:12.340 The Velocity is just going to be a vector attribute on the points. 07:12.720 --> 07:15.480 This attribute is also going to be part of the simulation. 07:15.481 --> 07:20.580 So the way to add this is to just click on either the simulation in or output. 07:21.080 --> 07:23.910 And then here in the sidebar, you can see that 07:23.911 --> 07:27.081 there is a simulation state panel that is new. 07:27.180 --> 07:33.420 And when I add another entry, this one pops up both from the input and the output. 07:33.980 --> 07:38.340 This is necessary because for the simulation step, everything that goes into 07:38.341 --> 07:41.096 the simulation output needs to pop up back 07:41.097 --> 07:43.840 out of the simulation input in the next frame. 07:43.841 --> 07:46.840 And then here for the socket type, I can just 07:46.841 --> 07:49.841 select a vector and call this one Velocity. 07:50.960 --> 07:53.680 Now there's two things that I want to do with the Velocity. 07:54.120 --> 07:57.310 First off, I want to keep it around for the next frame, 07:57.311 --> 08:00.720 so I'll pass it right from the input to the output. 08:01.340 --> 08:04.360 And then I want to use that Velocity to move the points. 08:04.840 --> 08:09.460 To make this independent of the frame rate, let's use a scale node and scale 08:09.461 --> 08:12.480 that with the time that has passed since the previous frame. 08:12.481 --> 08:17.020 And then connect that into the offset of the Set Position node. 08:17.500 --> 08:21.680 Now I can just set the initial Velocity for all the points in here. 08:22.160 --> 08:28.040 Let's say one unit per second in the X direction, and play it back like this. 08:28.580 --> 08:30.860 But so far, nothing interesting is happening. 08:31.980 --> 08:36.440 So let's make some changes to that Velocity, instead of just passing it through. 08:37.120 --> 08:39.360 Let's just add a force like Gravity. 08:40.220 --> 08:43.319 And the easiest way to do that is just use a Vector 08:43.320 --> 08:46.401 Input node, which will represent the Gravity. 08:46.780 --> 08:50.860 And then we can use that to modify the Velocity Vector for each frame. 08:51.320 --> 08:55.980 Again, to make this independent of the frame rate, I'll use a scale node. 08:57.320 --> 08:59.320 Scale that with the Delta Time. 09:00.440 --> 09:03.780 And then just simply add this to the Velocity. 09:05.420 --> 09:08.466 That way we have a change in the Velocity on 09:08.467 --> 09:12.501 each frame that is represented by this force. 09:12.820 --> 09:17.249 And if I replace this in the Simulation Output, you 09:17.250 --> 09:20.421 can see that this affects how the points are moving. 09:21.700 --> 09:25.720 So let's set this to something that kind of makes sense, in terms of Gravity. 09:28.020 --> 09:29.980 And that kind of looks right. 09:31.920 --> 09:36.480 Let's give this some Velocity in the Z direction as well. 09:37.980 --> 09:38.980 And there we go. 09:39.320 --> 09:43.260 So with a pretty small amount of nodes, we have a very, very basic setup already. 09:43.920 --> 09:45.840 So let's take this a few steps further. 09:46.820 --> 09:51.100 First off, I want to get a Random Initial Velocity for each point. 09:52.420 --> 09:55.400 So for that, I'll just use a Random Value node. 09:57.680 --> 10:00.260 Going into totally random directions. 10:03.320 --> 10:04.580 And then it looks like this. 10:05.480 --> 10:09.440 Another thing is I want to continuously add more points to the simulation. 10:10.960 --> 10:15.640 Right now we only get the initial state of the points through the simulation input. 10:16.380 --> 10:21.040 But this node is not the only way to pass in data into the simulation zone. 10:21.041 --> 10:26.100 Instead, you can also just pass it in directly, like this. 10:26.420 --> 10:28.500 I'll just create a Join Geometry node. 10:29.520 --> 10:35.421 And then join that together with the points that we get from the previous frame. 10:36.340 --> 10:41.960 And that way, we add a new set of points every single frame. 10:42.500 --> 10:46.720 Again, this data is just what we get from the previous frame. 10:46.721 --> 10:50.320 Or on the first frame, from the initial step. 10:51.120 --> 10:57.740 And then the data that we pass in directly is calculated for each individual frame. 10:58.300 --> 11:02.303 So if I play this back, you can see how we get a new 11:02.304 --> 11:05.500 set of points joined together every single frame. 11:06.080 --> 11:08.780 But the set of points is always the exact same. 11:09.120 --> 11:11.420 So let's change up the seed for every frame. 11:11.421 --> 11:15.960 That way we get a random distribution every time. 11:17.360 --> 11:21.041 But you can see that only the first initial 11:21.042 --> 11:24.581 points get a random velocity in the beginning. 11:25.440 --> 11:28.560 The other points just start out with a velocity of zero. 11:29.400 --> 11:33.502 The reason for that is that we only pass in the random 11:33.503 --> 11:37.880 value to the initial points in the simulation input. 11:38.680 --> 11:41.640 With the other points are joined into that 11:41.641 --> 11:44.280 same geometry, they don't have that attribute. 11:44.820 --> 11:48.660 So what we need to do is add a capture attribute node. 11:49.080 --> 11:50.360 Set this to vector. 11:50.960 --> 11:56.020 And then in here, capture this field as an attribute. 11:57.480 --> 12:00.860 And then add it together with the velocity of the other points. 12:03.580 --> 12:08.403 That way we combine the velocity attribute of the existing 12:08.404 --> 12:11.960 points with the new captured attribute of the new ones. 12:13.700 --> 12:14.960 And there we go. 12:17.060 --> 12:19.500 But let's make this simulation a bit more interesting. 12:20.340 --> 12:24.440 Right now we only have some initial conditions for the velocity. 12:24.441 --> 12:30.620 And then a single constant force acting on the particles to change that velocity. 12:31.240 --> 12:35.760 Let's add some simple turbulence to them by adding a noise texture node. 12:37.620 --> 12:39.860 Let's connect the viewer to take a look at that. 12:40.420 --> 12:43.540 And right now the noise texture gives values from zero to one. 12:43.840 --> 12:45.660 But we want to center that around zero. 12:46.580 --> 12:49.940 So let's use a map range node. 12:52.240 --> 12:53.240 Turn off the clamp. 12:53.241 --> 12:57.060 And put the from min to 0.5. 12:57.940 --> 13:00.960 And that way we get nice values centered around zero. 13:01.880 --> 13:04.880 Let's connect the two max to a value node. 13:06.640 --> 13:09.540 And then we can control the strength of this field. 13:10.040 --> 13:14.880 Now we can simply add this to the existing gravity force that we created before. 13:17.760 --> 13:20.340 And there we have some turbulence in our simulation. 13:21.100 --> 13:24.280 Change up the scale to make that a bit more clear. 13:27.060 --> 13:28.740 And crank up the strength. 13:30.220 --> 13:32.240 Now that's a bit more interesting already. 13:36.250 --> 13:40.010 Another thing I want to do is keep track of each particle's lifetime. 13:40.910 --> 13:44.370 Another thing I want to do is keep track of each particle's lifetime. 13:45.330 --> 13:49.990 So on top of the existing data that we're keeping track of in the simulation already. 13:49.991 --> 13:55.350 I'm going to also add another attribute called the age. 13:56.770 --> 13:59.970 Which is just going to be an integer to count up the frames. 14:01.370 --> 14:07.370 In the simulation we're simply going to add one to the age for each frame. 14:09.050 --> 14:13.110 But now by default we have a bit of a problem that's not super obvious at first. 14:13.111 --> 14:17.451 But this little dot inside of the diamond shape indicates 14:17.452 --> 14:20.810 that we have a single value for each of the particles. 14:22.370 --> 14:25.790 It's not actually an attribute, it's just a single value. 14:26.830 --> 14:29.650 That is another feature of simulation nodes. 14:30.150 --> 14:34.770 You don't necessarily need to operate on your geometry data with attributes. 14:35.290 --> 14:38.930 You can also just simulate individual values. 14:38.931 --> 14:45.350 So right now there's no way for the system to know that the age is supposed to be an 14:45.351 --> 14:49.090 attribute that is different for each point on this geometry. 14:49.730 --> 14:53.490 There's nothing that ties the age to the geometry right now. 14:53.770 --> 14:58.410 So to make sure that this happens we need to use another capture attribute node. 15:02.070 --> 15:04.830 And then pass the age through that. 15:06.070 --> 15:07.990 And now you can see that the dot appeared. 15:07.991 --> 15:13.210 And now you can see that the dot disappeared because now this is actually 15:13.211 --> 15:16.050 an attribute on the geometry in the simulation. 15:16.950 --> 15:19.974 We can keep track of the age for example by connecting 15:19.975 --> 15:22.830 the viewer and then looking at the spreadsheet editor. 15:24.390 --> 15:28.470 So the points that exist from the very beginning just count up more and more. 15:29.250 --> 15:32.553 But if I scroll down a bit you can see that the last points 15:32.554 --> 15:36.610 that are just newly created just have a single age of one. 15:36.611 --> 15:40.714 Now this age parameter we could for example use 15:40.715 --> 15:44.110 inside of a shader to create all sorts of effects. 15:44.390 --> 15:50.170 Like changing the color over the lifetime of a particle or inside of the simulation. 15:51.130 --> 15:53.452 For example to just delete some of the points 15:53.453 --> 15:56.411 when they reach the end of their lifetime. 15:56.490 --> 16:05.270 So let's create a delete geometry node and plug it in here to delete every point 16:05.271 --> 16:11.410 where the age is greater than let's say 20. 16:13.090 --> 16:16.370 So the particles stay alive for only 20 frames. 16:17.550 --> 16:19.330 And then they disappear. 16:20.610 --> 16:24.130 Now let's actually do something outside of the simulation itself. 16:24.131 --> 16:29.690 I want to make use of the velocity that we get for each particle. 16:30.810 --> 16:36.210 Now something that is important to note is that you should never use the velocity 16:36.211 --> 16:41.930 right out of the simulation zone itself to pass it out to some other node outside. 16:42.810 --> 16:46.139 Every connection that you make from inside a simulation 16:46.140 --> 16:49.670 zone to the outside will still be part of that simulation. 16:49.671 --> 16:54.533 So if you want to get data out of the simulation, the 16:54.534 --> 16:59.010 only way is through the simulation output such as this. 16:59.210 --> 17:02.150 Note that this is different for the other way around. 17:02.390 --> 17:08.850 As I shown before, you can pass in data from outside of the simulation zone just 17:08.851 --> 17:11.990 like this, but out only through the output. 17:12.730 --> 17:17.550 Let's use the length of the velocity vector to get the speed. 17:17.551 --> 17:26.130 Use a map range node to map it to a smaller range that we can visualize like this. 17:26.490 --> 17:29.510 But let's use a color ramp for a better visualization. 17:30.550 --> 17:33.910 Let's just change the colors to something like this. 17:35.750 --> 17:37.090 And there we go. 17:39.950 --> 17:44.050 Let's crank up this density by the way to get a lot more points. 17:44.950 --> 17:50.030 Because this is a very simple simulation still, the performance is still quite good. 17:50.430 --> 17:56.630 Now, because we are piping through the actual group input geometry into the 17:56.631 --> 17:59.153 simulation zone directly like this, we also 17:59.154 --> 18:03.050 have direct real time influence with the mesh. 18:03.950 --> 18:09.410 So if I just play the simulation and go into edit mode, you can see that I can 18:09.411 --> 18:13.270 change the mesh and see the updates in real time. 18:14.010 --> 18:22.030 So you can interact with the inputs like the mesh itself or all of the parameters 18:22.031 --> 18:26.670 that we are using in the setup in real time to affect how it behaves. 18:29.750 --> 18:31.630 Just as you would expect from a simulation. 18:32.870 --> 18:37.210 Let's add one more element to this simulation for some more interactivity. 18:37.211 --> 18:43.150 I want to add a separate force that is controlled with an empty object. 18:43.590 --> 18:49.450 For that I'll just drag the empty into the node editor to get an object info node. 18:50.010 --> 18:53.230 I'll set this to relative to get the relative location. 18:53.810 --> 18:57.672 And then do just a little bit of vector math to get 18:57.673 --> 19:01.290 a force that draws the particles towards this empty. 19:02.330 --> 19:09.891 Let's take the location of the empty and then subtract the position of the particles. 19:12.270 --> 19:18.730 Now this vector, let's just normalize it and also get its length. 19:20.890 --> 19:25.090 And I want it to be stronger the closer the particles are to the empty. 19:25.091 --> 19:30.877 So I'll use a divide node and divide one by the 19:30.878 --> 19:35.690 distance and use that to scale the direction vector. 19:37.530 --> 19:42.430 Then just another scale so that we get more control over the strength. 19:43.450 --> 19:46.570 And let's add this to the existing forces. 19:48.570 --> 19:54.890 Let's mute this link holding down control and alt, right click drag to get rid of 19:54.891 --> 19:58.790 the original forces so we can see better what this does. 20:00.610 --> 20:03.450 Right now it doesn't do much because it isn't strong enough. 20:03.850 --> 20:05.750 So let's turn that up. 20:07.430 --> 20:11.470 And now you can see how the particles are drawn to the empty. 20:14.960 --> 20:16.660 Let's crank that up even higher. 20:16.661 --> 20:22.780 Now we got a really strong effect like this. 20:24.060 --> 20:27.020 Let's turn these initial forces back up for the full picture. 20:27.940 --> 20:28.940 And there we go. 20:29.040 --> 20:33.340 This is our final basic custom particle simulation setup. 20:35.080 --> 20:38.288 Of course, this is just one of the infinite use 20:38.289 --> 20:41.201 cases of what you can do with simulation nodes. 20:42.040 --> 20:46.028 But I hope this gave you a bit of a solid idea of what 20:46.029 --> 20:49.140 things you can do with it and how to work with them. 20:50.100 --> 20:53.480 In case you do feel like some of the things in this video went a little bit 20:53.481 --> 20:57.600 over your head, I can recommend you to check out the Geometry Nodes from Scratch 20:57.601 --> 20:59.900 course that we have on the Blender Studio platform. 21:00.680 --> 21:04.160 In this course I go over everything that you need to know to understand how 21:04.161 --> 21:06.120 geometry nodes work and how to work with them. 21:06.740 --> 21:07.820 But that's it from me. 21:08.040 --> 21:10.040 Have a great rest of your day and goodbye.