 Svingen är här för att tala om Svällsjöck. Det är en nobbygd system för. Designning parametrik objekt, och jag vill starta varje sjunga short movie. Det första parten är med mig, och det är särskilt bra. Bara Nikitron. Det är det mängderlöpare behind produkt. Och jag vill lätt stressa att vi är inte animatörer. Vi är arkitekt. Det är just för att pläga runt med det här parameters av det. Jag har inte sett alldeles. Nice videos. everybody been showing is a bit. Shameful to show off this. Some of the pictures are very nice and some of the things are not so nice, but it's all about potential and not actually what we have done yet so much. These are the few manufactured objects like a table, a bar disk and this three to the things. So my name is Lin Sving. I'm a Swedish architect. I'm not a developer and if you would look at this red shock source code, I think it will after a while become sadly clear that it's not a professional quality Python atom, but it's a working atom and it can do some very nice things. I started in early 2014. I did these two nodes that you can see down to the left first. A simple thing to get a series of integral floating values and add the seed setting to the random to make it actually useful in the design context, because before it was continuously updating and so I started the whole thing by scratching and itch. And I just continued and it became a hobby of mine and I have to say I'm not using. This is any professional capacity. I haven't done anything professional with it, but some other people have been doing things and I will show some things. The development team is listed here. Alexander and Nikita started the whole thing. Augustin joined them and made some nodes and I joined them and I took over like some core parts and I continued developing. Seffi, the Algar McArdle, has been doing a lot of interesting work and Konstantin has done some very nice blender integration in certain parts. So it's a node based system to create geometry. We manipulate mesh geometry. We don't really work with curves. You can manipulate geometry by creating vector mathematics, simple mathematics and you can form this relationship between the vectors and vertices. You can create mesh geometry. So it's like it's an optimistic project and it's continuously growing. Right now the add-on download size is 800 kilobit, the zip file from GitHub. It's an incredibly large add-on for being in blender, I think. But it's also because it's a very, by default, it's a very modular system. Every node is a module and all the modules are basically independent, but can be reused in different ways. A visual programming language. The node group is evaluated as a directed ice-clicked graph, like all blender node groups are by default, which limits what people want to do. For example, we cannot do recursion, iteration and certain things. I have some work to do with abstractions for these things, but I haven't done anything with it. So why are architects interested in parametric design? This is excerpts from the parametric manifesto by Patrick Schumart, who is one of the main characters at Sahadid Architects. You might agree or disagree with what he says. I personally dislike him and lots of the work they do, but I think it's very interesting, the relationship between the computer and the tool and how you create things. What we can do is very deeply affected by how we do it. The tools available to control what we can do and what we can realize. So if you look at this example, you can basically see the mesh grids being built into the rebuilding. It's very easy to see, like NARV flow and other things, realizing the physical form. There's a very derelect relationship between the software and what it does and the aesthetics in the building in this case. So parametric designs in the Svanschok, this is by Delga McCardle. Sefi, his thinking term of causes, symptoms resolve automatically. If you think it can be done simpler, it probably can start. I heard a lot here from people today and the previous days that they had looked at the add-on, but very few people have felt like how to get started. And there's a serious problem with this system. So the work flows of this is that you have to realize it's a tool set. It's not a complete solution. There's no completeness in it. It's just you have to build everything by yourself. You give a use in tools and you can choose to use this as something simple. Just you want to place some geometry in a very specific metamatic way. You can do that and you can use that later as a mesh. You can do something powerful and create a complete building. It's extensible in a different way. The basic concept is you position mesh data and organize meshes. Currently there are 128 nodes. It's written in 100% Python using the PyNodes interface. We're doing a documentation. About 70% of the nodes are documented. You can see an example in the bottom part of the picture. The Blender artist thread has reached 80 pages and almost 130,000 views as of today. So it has generated enormous interest in this also why I'm here talking about it. So what can you do with this? You can play with geometry. This is a fish example by Delger McArdle. I choose it because Frank Geary once said, a fish is architecture but a cow is not. Another simple example is you can manipulate the vertex weights of a mesh which can be very useful in this. A very simple example we use the dot product that we remapped to a different range based on the vertex normal in this mesh. You can do with different addons or different things. This is a generic tool set that you can do many such things with and you can continue doing this. You can use it for simple things like this in many different ways. We have an extensible scripting system. We have two different script nodes where you can input your own simple code to generate things. In this case there is extremely simple generating a grid. I won't go through the code. It's very simple for anybody that knows Python and you don't have to deal with these things. It's to show that if you want to do your own node or you want to generate your own geometry you can very simply input this and parameterize it and you get the results. This is a very simple flower petal done with sinus curves. It's not hard to model this in other ways. This is a very simple demonstration of how to create variation. Everything in Svertock is a list. If you input a list of numbers into the number of petals here you get a series of different flowers and you can position them in this case. We use a plane as a generator to generate a series of points to distribute the flowers in space. Or you can do something extremely advanced as Philippe Guimiot, Dormon architecture.com. This is a set up of 350 nodes. I can say that this clearly highlights some problem of the current approach. We don't have node groups, we lack many abstractions. There's not a lot of unneeded repetition here. But many of the we have different techniques in ways now to simplify this. My intention was to show how simple this layout could look compared to this but I haven't had time to disassemble it and reassemble it in a good way this weekend. I've been to busy drinking beer. I want to show some simple files. You can simply animate this fish in a very basic way and it's not a useful animation perhaps in the sense that you can manipulate it in a good way. As we said we're not animators but you can very simply parameterize and show the parameters. I won't really go through this example right now because it takes some time and it's not really perhaps whatever. I will do a very simple basic demo a little bit later just to show this. This is a simple amount of nodes basically for doing a very simple thing and you get something like this. And this is the complex example the building shown before. It's a totally parametric object. There's no external geometry. Everything is generated from numbers in different steps and solidified offset and many different operations are applied to it. And here you have the settings of the node tree that we exposed in the user interface. So if for example this is the floor height parameter it's now before 3.1. Let's set it to 1.5 so we have a midget building and it updates this very quickly and the updates are not quick enough to perhaps to drag these numbers on my computer perhaps this one is quicker but every parameter that you can design here you can animate or you can do other things with it. Of course this I want to show you this node graph. Total craziness. I mean this is an unacceptable amount of complexity perhaps with signing a building if you look at it firstly but it's all made in like it's very repetitive the operations applied and if we had node groups it could not be done much simpler and I won't say that this building perhaps is like how you should design a building or anything like this but it's a possible way to design a building or design geometry use these controls and it's a horrible, horrible mess in many ways this model and one of the problems we have that is every change you do in a setup like this takes a lot of time and if you have used it recently basically until this week there's a new version available when I solve some of these problems and I have time now to change fix things but there are for every change when he did this he have to wait like one second for the node tree to update every time I don't know how he could do it like this would be too annoying to me but it resolves now finally and we also expose a series of settings like you can lock it from being animated we can show to display it in the OpenGL preview or you can bake it if you want as a mesh object so this is a very simple object I want to show you some basic techniques Sveljschok is based on that we have nodes to produce data and we have output nodes that you can use to visualize the data or produce a functional mesh object that works with almost all modifiers I think the skin modifier is the only one that doesn't really work so in this case we have taken four circles lifted them and created a cylinder and by changing the parameters here of the width the diameter of the circle we can do whatever we want with it and we position the circles by creating a list of numbers that we input into a vector we get a series of vectors that are locations for matrix and we display the vertices using a matrix operation and from this simple one we can then interpolate another surface that this is using cubic interpolation and you have the same four control curves that you can then choose what will happen with the object and there you have the mesh object instead it's a proper blender mesh object that you can then continue to work on and keep modifiable and we can construct many different types of geometry and you can choose this won't be anything this won't slow anything down until you change the parameter when it has to recalculate of course if you edit the mesh object and you recalculate it you provide your original your edits but it's unavoidable in this kind of context and here we have a very simple I mean this is extremely in this case simple setup it should be even simpler of course we have a series of interpolation objects being created from the vertices that we change the direction of the lists and a lot of things and the main concept of course is that everything is a list of lists and constraints numbers so you should be able to modify create a number of objects in the same sequence all the time this is somewhat working sometimes and sometimes not working depending on the age of the specific node so what we need to do we need a better structure it was started as a hack, a very optimistic hack to get things done and we got things done but it has grown perhaps too much for its own good and the number of code and the number so it's becoming harder to try to change it and as I said earlier we are not developers especially not me but I have some ideas about how it should be working and we are very far from that but even today you can do very very powerful things with it I think and standing here at the conference and looking at what all the other people are doing I feel because I'm a blender I'm a user of blender, I don't know so much so I feel that many of the things that I've shown here can be done of course in better and like more standard ways so we also need, we need carve objects, we need text objects we need real, to work with real objects and similar things to do better of learning integration of course even today you can already achieve this by using these scripts you plug into the system for example you could do a dancing node layout which is kind of useless but you can do it another point I would like to raise is that there are now like a series of different node systems that kind of could work together but they don't and I think that for the future it would be very nice if you could unify these systems perhaps not our system is not supposed to be the base we might even be like a hindrance in some ways but I really think that there is a great potential in a node based design where you can manipulate the parametric objects and the parameters of what you want to design and keep it flexible so I want to point you to these two web pages the Github and the Nikkejons Russian thread where you can find a series of lessons in Russians that are actually quite usable with Google Translate if you don't speak Russian yet and there's also the Blender Artistress we are always welcome to ask questions and come with input or any requests or anything of course we can't promise anything for more technical issues of course we prefer to use the Github issue system and everybody is welcome to participate or look at code and we welcome any help with either design or any help basically let's stop there En amazing example of the power of our Python APR so I'm still wondering if Campbell is not here the guy is calling probably but did you meet with him? No I haven't spoken with Campbell he could look at this he really likes these kind of systems and he could probably advise you on some speed up or he might get inspired in optimizing some parts of the API especially the parts where you have to handle lots of points or mesh data which is possible to make much faster any quick questions from the audience we have a few minutes we have a lot of time because there will be a half hour break so anyone has a question about this Adam you are an architect can I answering my own question I can make my own Python node right? Yes of course Is there also a script node that I don't have to make it in? Yes and everything is animatable 3 questions sorry so we have these script nodes here we can load let's take an for a very advanced example in a way We could have done a PyNode workshop for we should do more workshops anywhere in the microphone because it's little Python script little pieces of code and they convert to nodes and then you have a visible programming language for your pleasure so everybody can make little script convert it to a node and then you start noodling around that's how it works mind blown so I'll try to do that I'm very impressed I wasn't here since the beginning where can I get this? You can get it at the GitHub address or you can get it at the GitHub I should mention also that the version I'm showing today is the experimental version available on GitHub if you want we have a different series and this is the version I was showing today that has some node groups that are kind of working and other new features I have a question for me this thing is just wonderful and I think it was really missing in Blender and I'm teaching in the School of Architecture and they start to teach another software I will not give the name but I think you know which one and is there a stable branch that we could use already to teach it or are we I would say that I try to keep it as stable as possible but we are a couple of developers and not everybody has a shared same idea about like stable and how everything should work and I think that partly it should be redone and that will probably break some things but I think you can use it it's working in a lot of ways and we cannot develop it without people using it and trying to find new use cases so somebody has to take a leap and give us feedback and we can improve it but I would say honestly that it's not comparable to the other software but you can do amazing things The other software is not used completely but today it's full presentation and I think here we have enough for teaching what is parametric design and this is what is very important for me Thank you One more, Madame Claude I think it's happening We have a short break Thank you for your talk In 13 minutes we are going to see Monica with her Bio Blender project