@Benjamin Bass Currently it's not possible to set this up automatically unfortunately. This is a known issue and the solution needs to be properly designed still, as we want to have a better solution than how it is solved with shader nodes currently.

@Simon Thommes I was wondering about this too. As a workaround, could you do something like plug Object Info -> Location into your 'seed' of the random value node? Then if you were to duplicate the objects in various locations, they'd sort of have random starting values? I played with it a bit, and for a second I thought it worked, but my results were inconsistent.

@Tyler Bjarke The object info node still needs an object to explicitly be passed into it, so that would need to be set each time the node is used, which defeats the purpose. In Blender 3.4 (allpha) there is a Self Object node, which solves that issue. However proper randomization still needs to be done. The same approach that is used in shader nodes is purposefully not added to geometry nodes as it is quite flawed. The current plan is to enable reading out  custom object properties in geometry nodes an then adding a feature that randomizes a property when duplicating the object. That would allow controlling that seed value explicitly and is more stable.

@Mahesh Deshpande Yes, totally. You can use the Collection Info node to receive a collection as a geometry instance.

@Alexandros Hadjilaou Currently that is not possible natively. In the future there will be a node for that. But right now you can make a nodegroup that scatters points on a surface the scale of the bounding box. Move the points around so that they cover the volume of the bounding box and then delete all points outside of the mesh.

@Sorcière and grease pencil !!!

@schmutzer1 I'm not exactly sure what that means. For me the downloads seem to be working perfectly fine. The direct links work fine even without being logged in. You're saying ' tagged as moved or disappeared immediately after download', if it is after the download, that sounds to me like it would be something on your end.

@Simon Thommes Hello Simon, thank you for your support. You're completely right. It was a browser specific issue with Firefox. Fortunately it worked fine with Microsoft Edge. Congratulations again for your this great course and I wish you all a nice summer at Blender Studio.

@Katie Korwal correction 16:42

@Katie Korwal Float is a technical term for a number with a floating point. It's the same as what is just called Value in shader nodes. The reason I'm changing this here is that by default the input was a vector, but I only want to control it with a single value. Blender always implicitly converts between different input types when possible. So the float input that I give in is then converted to a vector by using it for every component of the vector.

@Michael Gembicki For me, opening the file with Blender 3.3 everything is working just fine. The seed being an input field sounds like the input was toggled to an attribute with the button next to the field. But I just downloaded the exact file in the description and opened it with Blender 3.3 and it's all fine. Can you check that again to make sure?

@Simon Thommes Yeah, redownloaded example and same issues.  This is my blend file saved from 3.3 https://drive.google.com/file/d/158sFbzycay3jrPzwDRaUAYPYIJ8MOP0T/view?usp=sharing

@Simon Thommes I tried it on a vanilla version of 3.3 and it worked just fine.  It must be one of my, way too many, addons causing it. Thanks for the response. I'm enjoying the course.

@Michael Gembicki Ah okay, yes that makes sense then. It seems strange to me how an addon would mess with this file so specifically, but that can happen. At least it doesn't seem to mess with the file itself. The one you shared also works just fine for me on vanilla Blender

@Huiquan Lu I'm not sure if I understand correctly. If you are referring to the group input nodes: There can be however many group input nodes you want and there is no difference between using a single one or multiple. I like to split it up into multiple nodes to organize the nodetree in a more readable way. The group output can only be a single node, as otherwise that can lead to ambiguity.

@Simon Thommes Thanks Simon! Yes, that was what I wanted to know. as I watched to the end of the video, I started to figure out why you did those separated input nodes.

@Mark Hannen This is a setting that can be enabled in the preferences. Extra Shading Pie Menu Items under Keymap.

@Dan Caldwell I am using Blender 3.4.0 by the way, has something changed between versions maybe?

@Dan Caldwell That is unfortunately just how it works. The issue is that in the material dropdown you edit the material slots of the original data. But while evaluating the geometry nodes there are additional material slots added for new geometry. Those can only be assigned with geometry nodes in turn. What I usually do is assign it both with geometry nodes and to the material slots so I have a context to select and edit the material.

I agree that this is not ideal. But the behaviour is at least consistent and the usability will get better hopefully.

@Joshua Houston Is this with the latest Blender release? If so it would be useful if you could identify at what step the crash happens and report it as a bug, so the developers can take a look at it. In Blender under Help - Report a Bug

@Alex Dumproff > A mouse hover over the input socket doesn't tell me that. It does actually. Whenever a socket provides the same value for each element it will actually show that value when you hover over the socket. Otherwise, when the value differs for each element the tooltip with show it as a field.

The random value node will usually give you differing values for each element. Essentially it always gives you a random value per Seed AND per ID. If nothing is connected to the ID that means that is falls back to the ID attribute which is stable when you delete points or the element's index, if the ID attribute doesn't exist. You can think of the ID input as being a second seed input. The only way to receive the same random value for all elements is to plug constant values that are the same for all elements into both the ID and the Seed input.

@Nico Ma For me that still works the same in Blender 3.6. Note that it is crucial that you set the link type of the material slot to Object as shown in the video.

@Simon Thommes Yes, I set the link type to object. It just wont work. The most weird thing is,  when I render a test scene, the rocks have a grey color. But nowhere in my project, neither in the color ramp of the shading tab of the material, nor the viewport display color, nor anywhere I can think of, is a grey color left. I now even deleted all the colors from my project in the outliner blender files tab. In the viewport shading: material preview no color is applied at all, but when rendering the rendered image (the rocks) is colored grey? Ive been working with blender full time for 3 weeks now, and I do not understand this behaviour at all.

Edit: After some testing, in another project file with a similar setup, I was able to assign the color without FULLY using the Set Material Node in this way. In the Material Properties tab I set the link type to Object and linked my material. First, nothing happend, but after just inserting a Set Material Node right before the group output the color was applied. But the Set Material Node had no material selected! When I now select a material with a different color in the node, all Objects that share this node network get this color, except for the one object where I set the material in the properties tab via object link.  So to me this looks just wrong. Because used this way, I need the setmaterial node, even tho I only want to set the material in the properties tab. Obviously something is wrong, either with my blender, or with the version itself.  Most of all, it very much looks like you were able to perform this coloring via the material poperties tab just fine, since not a single set material node is to be found in your example rock_scattering file.

@Nico Ma This is just how materials work in Blender currently. If you link the material slot to the object then also only that object will use that material assignment. If you link it to the data, you run into the problem that the data is not actually part of the actual data of the object, because it is procedurally generated, only the node-tree itself can assign the material in that case.

I agree that the UX of this could be improved, but it is consistent with how it has been working so far.

@Bryan Green It helps with a cleaner node-tree, as you don't get noodles running over the entirety of the tree and it logically groups the inputs where they are actually being used.