You need to login to view this content
You can join Blender Studio for €11.50/month and get access to all of our training & film content instantly!
Theory - Attribute Propagation
16 comments
i think this video might be more important than just the subject matter. I struggled with this topic for a long time and seeing it explained this way brings alot of concepts home. Just great.
Hi there! I have a few questions I wanted to ask, since trying to figure this out on my own does not work. OK, so I have been trying to 'capture data' in all the ways I can find, but it does not work as I expect. For example; I can use a mesh line with a certain amount of points (say 100), and then apply position data to this line in different ways, so that it for example creates a sine-curve. So far so good. But, what I would like to do now is to 'transfer' this data set to other objects in a 'controlled' way. Sort of like a driver or something like that. Say I want a cube to follow the same path as the already made sinus-curve, how can I do this? And if there is more than one way, what do they look like?
As a second wish (suggestion) from me; I would really, really like it to be small examples on (if possible) every node there is (in the manual that is), so that we, the users, can build on from there. The simpler the example, the better it is. I think the community will benefit greatly from this, because it will create a lot of new examples of what can be done--much in the same way as the Blender Studio staff is doing now (doing big projects to push Blender further), but by the community. I mean, I watch a lot of community made videos, and learn a lot from that; but even the smartest ones of them struggle with some of the concepts (of which there are many in Blender), and that is why I think it is very important to start breaking Geometry Nodes down to the smallest pieces possible as a way for us (the users) to really get an understanding of how things work inside Blender, and start being creative with it.
I'd would like to discuss this further also; if anyone thinks this is a good or bad idea from me.
I work as a teacher on a high school in Norway. Subjects are creativity in all forms, using new software and programming as creative tools. So we use Blender there, but so far we keep it to the modelling and rendering part, but I really would like GN to be easier to understand for everyone so that we can start using that as well here at our school. Keep up the good work! :) PS: Your instructions are great, although somewhat demanding to follow because they are also very dense on information, which is a good thing also ... :)
@Knut Einar Skjaer Hi Knut!
Regarding you first question: When you are trying transfer information from one geometry to another and they are not related to one another in the node-tree, you have to use the sampling nodes. There are different ways to sample geometry, e.g. proximity, raycasting, UVs. To model something like what you are describing it would be best to use a curve geometry, rather than a mesh and then use the Sample Curve node to get transfer the information.
To your second point: All nodes are already described with a certain amount of detail in the manual. Only some of them contain examples, but creating and maintaining examples for every single node there is is a lot of work, which is better for developers to spend on improving the actual system itself. That said, Blender is an open source project and contributions by members of the community to Blender itself, including the documentation, are more than welcome.
Lastly, thanks for the kind words! I'm aware that my courses are quite dense in information. It's not possible to cater to every person's individual learning needs in an online setting like this, so my focus was to be concise and complete in a way that it can be rewatched with specific points in mind as the viewer gets their own experience with Geometry Nodes.
In terms of difficulty to get into Geometry Nodes: I think we're already striking a quite good balance between complexity and accessibility. I know plenty of people that would describe themselves as not very technical but are successfully building node systems for their needs. Of course we can still do better, the idea is to ship Blender with more and more pre-built node groups like we did for the grooming workflow that should take away a lot of the burden of complex setups from the user.
@Simon Thommes Thank you very much for taking time to answer my rambling! :) I fully understand what you are saying about the difficulty of having/making examples covering every node there is; but, I still think it would be a good idea to work on that. Just to give a 'hint' of what they are meant to do at least. The examples might be best if they cover more than one node at the time, though, more or less like what you are doing in your videos (but shorter and less complex in what they achieve).
I do slowly learn how this -- geometry nodes -- works as I work on this ... but I also see how much time I spend trying to figure out what I think should have been simpler ... And I really do like to play with this, don't get me wrong, but it lacks a good bit of better documentation ... or more clarifying tutorials from people like you and other very competent people. If the documentation had been somewhat more detailed and clear I am sure there had been a lot more of good examples out there for anyone to enjoy and build upon. I have been in love with Blender since 2004, so please take this as an attempt of being constructive and helpful, and also just ignore this if it is in any way taken as something else. Keep up your very good work, and I would really like to see some more tutorials from you in the future :)
Kind regards, Knut Einar :)
PS: Thanks for the tip about Sample Curve tip. I was using Sample Index in a test after I wrote my questions to you, and that also worked pretty well for what I was trying to achieve ... :)
I am relatively new to the industry, Blender and Nodes- I found this series of videos to provide the foundation I did not realize I needed coupled with the inspiration to strive to learn much more! The accompanying support files are terrific. Well presented- Rock on!
@De Etta L Ewing Thanks for the kind words, I'm glad that this is so useful for you!
Do I see it right that a "Capture Attribute" node work in conjunction with a "Realize Instances" node creates an "ID attribute" for every vertex of the realized instances so that we later can address vertices created from different instances separately? And the "Random Value" node creates a separate vector for every group of vertices that receive the same instance ID within the "Realize Instances" node?
@Ingmar Franz yes, you can totally do that! That's exactly how you can use the different domains to your advantage,
I find that "Anonymous Attributes" should be better called "Static Atrributes" since for example the "Position" attribute also has a rather general naming. The "Capture Attribute" node creates a snapshot of a changeable "Regular Attribute" (which I would therefore rather refer to as a "Dynamic Attribute") and by doing so turns it into a "Static Attribute". It further looks to me that the "Anonymous Attributes" like the "Top" attribute of a cylinder are evaluated internally directly with the creation of mesh primitive (and therefore bound to it) whereas the "Position" attribute gets evaluated externally with a separate node which therefore can be connected to every node that refers to geometry.
@Ingmar Franz I do get your point and how you understand the concept is totally correct.
The reason why they are not called 'static' is because they do change in terms of interpolation. That is part of why they are so useful. Being able to use the same attribute after a topology change/converting to a different geometry type. Anonymous means that the attribute cannot be called by anything else than the exact socket where it is created. So by that it cannot explicitly be changed.
If it helps you to think about them as static that is perfectly fine though, as the behavior you're describing is part of their key feature. We had a geometry nodes workshop recently and part of that were some UI topics, which includes creating a stronger visual connection between a field and where it is evaluated.
Do you by any chance, have a tutorial on how to make the doodle clouds? I've tried replicating your example, but after the Fill Curve node, the curves completely dissapear.
@Anca Diaconu Hm, that's odd. The file is here you can compare it with that. Other than that, I just made a quickly cut mini tutorial for twitter.
Hello teacher, I want to ask about the difference between value and float, these two definitions have been giving me a headache
@mark mi Hey Mark, this is a bit of an unfortunate legacy naming. Float is a more accurate description of what the actual datatype is, while value is more loose. The reason why in previous node systems this didn't matter so much and it was simply called value was that Blender handles the datatypes on the backend while the user doesn't need to be concerned with the exact type. With geometry nodes the user is much closer to the actual data-structures and it matters a lot more. So there it is useful to differentiate.
@西蒙·托姆斯 Thank you very much
This one video helped demystify attributes more than any other I've seen so far. The fact that the Geometry Nodes development swung wildly back and forth between workflows didn't help, but glad to see that a sensible middle ground was found. Love this series! Thank you, Simon!
Join to comment publicly.