@Erindale , another tip 😁: when you were checking to see if any selection were too small you didn’t have to separate the mesh (which is some i try to avoid. [messes up my indices, attributes, extra compute time, etc]). what you could do is the less well known node family called the "face groups to boundary edges" and "boundary edges to face groups". You first use the selection as a face group which you convert to boundary edges and then you convert it back; this gives you an ID per selection island, which you could usr with the accumulate field etc.
Thanks Erindale, Your videos are really helping me do things impossible to achieve with built in modifiers. I hope you make a series about UV mapping especially mapping a "curve to mesh" geometry which uses a cyclic curve, I've been looking all around the web for a working solution but all of them work only in simple cases.
Try searching for it on my discord. It’s definitely something that’s been solved by a few of us. Fundamentally you want to create geometry that matches the topology in terms of face corners and then just transfer by index from one to the other.
@@Erindale I've tried this approach but the flat none cyclic curve results in 1 extra column of mesh indices compared to cyclic mesh. I'll definitely join the community and look for a solution. See you there.
@@Erindale I've tried this approach but the flat none cyclic curve results in 1 extra column of mesh indices compared to cyclic mesh. I'll definitely join the community and look for a solution. See you there.
Really great video. Just to note that for the 'expand selection' part, the Blur Attribute also does the job when you put it through a Vector Math (Ceiling) node, so no need for the Repeat Zone. (or you can multiply instead of ceiling for a more circular expansion).
grand. selections are an area where I seem to invent new struggles every time. this is super valuable and I'm sure I will trigger a few hundred plays in the near future. thanks for sharing!
Correction: Accumulate Field node's "Total" output does not tell you how many elemtns have been counted, it returns the sum total of whatever values it has counted up. When the Value is 1 as in mine, that means it added 1 per element which is why I worded it that way.
Your Boolean NOT is going to be the inverted selection but you still need to evaluate the mask everywhere. The selection sockets on Set Position, Store Named Attribute etc do limit the computation of the node operation just to the selection but, unfortunately, you can't get around evaluating where the mask actually is.
Thank you! I think that GN is a amazing tool, but people at Blender foundation should start thinking about a more intuitive way of working with data. In my opinion lists should be the way to go. For example, Grasshopper for Rhino is just way easier to manipulate your data.
It’s a different methodology and both Houdini and grasshopper (which is essentially a pythonic data flow) were considered. Once you’re fluent with GN you’ll find that a lot of things work a lot better and more intuitively because you’re specifically not needing to work with the data yourself. It’s handled inherently in the workflow
Hey, @@Erindale. Thanks for your feedback! Yeah, I think I'm trying to figure out GN logics with a pythonic dataflow in my mind, as you said. I think it make things a little harder, but I'll figure it out with time. By the way, your tutorials are helping a lot! Thanks!
Another superb tutorial, thank you. Using the Normal method, with either a Direction or Dot Product comparison, I can select lines (/faces) of latitude on a sphere or torus - but I can't figure out how to adapt that to select lines of longitude. Any ideas?
Take the normal, multiply by (1,1,0) to negate the difference in vertical direction and then normalize so each band has the same vector on it. The you can compare against a horizontal vector to make the selection
Great tutorial as always, thank you for that. I have a question - why does the float to integer node even exist when the exact same operations can be done with the regular math node? Is it more lightweight, or is it something else?
Integers are not susceptible to float rounding errors. It's easy to think of numbers as numbers like how we see them but computers have to do some trickery to actually store them. That's why on the compare node when it's set to float, equals, there's an epsilon value. Asking a float when it equals exactly itself will often fail due to some small decimal part being different. Integers also take less memory to store so it can be an efficiency gain, especially on very dense geometry
Thank you! Yeah it took me some time to get fluent with geo nodes and I realised that it's really all down to being confident in what selection you're making
@@Erindale I still watch all your videos in hopes to get that aha moment some day. Thank you for putting out quality contents for free. I'll definitely be purchasing your Simulation nodes course once I get the hang of this.
@@hereb4theendthis stuff fits squarely into networked muscle memory, so what you'll want to do is go from the top down. Understand the broadest purpose of geometry nodes, and then understand the broadest purpose of the different categories of them, and then look at a bunch of things people have done with them, focusing on the results. Then without even knowing exactly what all the nodes are, you have a framework for what you have and haven't seen done, and where you'll most likely find the components you're looking for to get the effects you want, along with the names of some of the ones people use all the time like instance on points and join geometry. Then it's a matter of drilling some stuff into your head over and over until you might not realize how much progress you made, so if you get imposter syndrome or feel like you're not doing anything, it might help to keep objective progress reports. I sometimes feel like I haven't done a lot, but then I remember when I was even clunkier, and all the surrounding circumstances I tend to not factor in as much as I should, and I build a little confidence and get my head back in gear
The tutorial was amazing as always, but finding out the most common and still yet least spoken issue and making a tutorial for it was even more brilliant! great job and thanks a lot dear Erin :)
Is there a way to use multiple vertex groups so that I can cycle through defined sets of geometry? (I’m trying to adapt a setup that just breaks a mesh up based on each face to use larger pieces that are more representative of how the object might come apart) - I’ve knife projected the pattern I want onto it so creating the groups might be a bit laborious, but at least I’ll get the result I intend 😅
I would name your vertex groups with integers like 0, 1, 2, 3 etc. Something very simple like that. Then have a node group that has an integer on the front. Inside, plug that integer into a value to string node, then the string into a named attribute node. That'll let you cycle through the groups. If you want a more complex name then you could use an index switch node and plug multiple different strings into it and cycle through that way
Wished I watched this waaaaaay sooner! really love this, the way you explained and went through this was great -- didn't even feel like a 1 hour video. Been following your other tutorials as well and have been trying to learn bit by bit through different use cases but really glad to have come across this as a general knowledge tutorial which I feel is incredibly helpful also once you start moving into more specific use cases where tutorials might not be available for.
Awesome tutorial. This is something what I was looking for. Giving a general information that can be used for problem solving instead just solving one problem. I'd like to see more tutorials like this in the future. This helps a lof for people like me who wants a deeper and behind the screen understanding for the nodes and math in general. I'd like to know more about the topology nodes as well. I'd like to know how you switched the connected socket from 'x' to 'z' so fast at 23:24.
Thank you so much for this. Both for excellently bridging the gap between a lot of other geometry node tutorials and Being Actually Useful, and also for using the word 'indices' repeatedly! Seriously, I had a use-case I'd been trying to get geometry nodes to work for me on for a while and every time I tried I watched some introductory tutorial, went to have a play, and got nowhere and gave up - and it was always down to being able to actually pick apart the mesh to do the work on the right edges. I've been going back to it while watching this and it's already 90% of the way to where I need it.
That’s great to hear! Thank you so much. This is one of those areas that’s just not sexy so a lot of people don’t want to talk about it but it’s got to be one of the most important things to know when working procedurally. Well done getting your project on track!
Warning about selection using Indices: The devs warn not to do this. Their backend is whack, and when they make edits to nodes, it could result in completely different indexing. I have an 18,000 node model that uses selection by indexing exclusively... and Each version of blender breaks at least 3 nodes... The one that pretty much made me quit using blender: Split Edges worked perfectly... and Hans went and redid it so that it *RANDOMLY* reindexes the *entire* model for no reason. When I asked him not to... he added notes to the documentation not to use indexing as selection. -_- Purposefully breaking nodes just rubbed me the wrong way to the point of losing all motivation to work on any Blender related project. Years of work... broken by someone else... on purpose. (He didn't purposefully break my node, but he did purposefully change the node to explicitly rearrange indexes, which breaks any nodegroup that uses index as selection... and refused to care that it broke years of work.)
Indices in Blender have always been a bit hit and miss. If you ever did anything with python building meshes outside geometry nodes you'll have run into the same problems.
wahooo, I have zero knowledge in programming. But this course is just master class. Thanks a lot ! This base will let us learn geometry node. Great explanation, considering I'm French native speaking. again Thanks !!
Great Tutorial. Thank you! To be honest, your tutorials keep pushing me to the limits, so I always need to watch them at least two times. Once I get it, it's really satisfying...
The vast majority are the same, and all should still work. There may be some optimisations you can do with new nodes though. A lot of curve topology nodes and little hacks with the accumulate field node for example
Thanks for the video! Ive been trying to work out if you can take a selection and transform it, ie move it, scale it and deform it. I'm aware all a selection is is a a 'True' or '1' against a vertex (or face etc) . So moving a selection means moving the value to the next vertex in a given direction. I know that if you want to move a selection, you're best to do it when your setting up the selection (i.e. if you're creating a selection with the raycast node). But it would be very useful to be able to take a defined selection and transform it. The only example of this type of transform I can see is the expand/contract operation covered in the video (others like higgsas have also implemented) It's hurting my brain trying to work out if its possible to do this, as your moving values across a mesh rather than moving the mesh itself! 🤯🤯
With mesh topology nodes you should be able to check if the neighbour is selected or not and change the current selection state based on that. The issue is that you’re going to need to do a bunch of selection-specific weights to make sure you’re checking against the correct neighbours for that specific selection on that specific topology… good luck 🥹
@Erindale that sounds horrific. haha! I had a vision of creating selections, using Boolean operators to 'shape build' the selections, then move / deform the built selections as a final step (tweak). Looks like this final step is going to be limited to expand/contract (I'm not mentally capable of implementing your suggestion). Appreciate the response! 🤝👌
Hey Erindale! Great tutorial and a lot of info. Thank you very much! I'm currently trying to make nodes for converting text objects to Neon Geometry. And stuck on the Neon ends. Which have to be bent backward from the letter plane. Main issue after i;ve deleted segment by using Endpoint Segment node for defining that segment. I have to move the ends backward. That is not a problem too with Set Position + Endpoint Segment nodes. But I also need to move that ends straight under the next points at the curve to make a straight angle between ends and letter plane. And I can't catch that, how do I get the location of the next points and use it for the positions of the ends. I'm a beginner here and might I've chosen the wrong way?
One potentially easy option would be to convert to mesh without a profile so you have an edge, use the extrude mesh node to extrude back as far as needed and then convert to curve to continue. You could even use the filet curve node to round that corner to the extrusion
Thank you Erindale! Many ways, and some of them are not so oblivious at my level. Looks like I have to learn the basics better. Then I will see.@@Erindale
Hey Erindale! Great tutorial and I learned a lot. Thank you very much! I have a question regarding selecting the UV Seam. I'm trying to select the UV seam to use it to split the edges in GN. My setup is basically getting the UVmap attribute from the group input and then run it through an evaluate on domain node with data type as vector and domain as points and then send it to a compare node and compare the UVmap original vector with the new evaluated one with a compare node with the Element-wise mode and Not Equal operation and capture the result from the compare node with a capture attribute node set to Boolean Data Type and Face Corner as Domain and use it as the selection for the Split Edges node. So far it seems to work in selecting the seam if the seam is going through a full loop cut but if it’s a partial loop cut it seems like it doesn’t select the final edge of the seam and it stays welded. I tried many different ways but I couldn’t get to select that last edge. Any idea how to get around this issue? Thanks in advance.
This is kind of a hack and it won’t be as performant as purely doing it via data, however, if you capture the original positions, split all the edges, set the positions to the UV map locations, merge by distance, and then set the position back to that original position, you’ll have the split edges where the seams have split the UV islands
Thank you, thank you, thank you! I am leaning Geo Nodes and up until 40th min I could understand and replicate almost everything from memory, but after that it become much much harder for my level - anyway will go systematically with you other tutorials and resources and try again in few weeks.
a really concise roundup here thanks... i think it mostly just clicked for me now prob the last tutorial i need (he says) the docs on these are so bad (literally almost 0 examples, just a small description, if you're lucky a "hint"), so this is great
excellent as always - but what is the difference between a 'float to integer' node and math node set to floor? don't they both just truncate the thing (by which I think I mean chop off the bits after the decimal point)?
The difference is how the data is literally stored in memory. Floats have floating point precision error so it just depends on your requirements. Functionally they’re so similar that if often doesn’t matter. Storing as integers will be more optimised
Make sure that you are connecting your viewer automatically with the hotkey so that it makes the right kind of socket. You see when I first drop on the white noise node that it shows black because it had been previewing an integer before. Then I recreate the connections on the float (grey socket) and then click again to get to the color (yellow socket)
Hey, great topic and deep dive. I noticed the pile of books behind you and since im a huge fan of 7 habits of highly effective people, i wondered what the rest of the books are.
In this stack I have Building a Second Brain by Tiago Forte which I definitely recommend. Also Deep Work and So Good They Can't Ignore You (by Cal Newport), The Freelance Manifesto (Joey Korenman), Eat That Frog (Brian Tracy), and Getting Things Done (David Allen)
Great tutorial, this is not a very intuitive way to work but with practice, I can see myself using this. I really hope that GN developers make things easier in the future !
@Erindale at least when I used it, if you compare greater than 0 and you converted it to a float, I haven't maxed out before blender slowed down significantly. and yes, the iteration matches the steps perfectly. it uses topology to smooth, so smoothing depends on topology density and always follows the topology
Great video. Figuring out certain things about geometry nodes -- selections, how to conceptualize working on fields and operating on them with some precision -- has been the biggest hurdle so far. This helps.
It's hard going but I think I'm starting to get the hang of the concepts. I like the way it teaches the fundamentals, not just press this, press that but explains the thinking behind it. But I'm a long way from being able to do it on my own, maybe one day!
Gold. First, I'm so happy that I've managed to figure out a lot by myself, but there are things that couldn't be possible without you and other authors on YT. Mostly because of terminology. The math itself is not that hard, but learning the names of math ops, as a kid, in Croatian still gives me headaches when learning GN watching the tuts English. It's a lot easier for me to build node systems for terrains because I am not limited by numbers that much and that's where my brain has all the artistic freedom it needs, and, basically, I often make tools from my experiments and happy accidents. Anyway, for the 100th time - thank you.
I'm going through this again and thinking how you, as an educator, would have a special place in the world of architecture. Maybe even as an architect. I know you had that sort of content in the past, but maybe getting more specialised in that area is even better than focusing on games/cg market/industry. Just a thought.
