# Games Have Never Simulated Clothing Like This Before ## Метаданные - **Канал:** Two Minute Papers - **YouTube:** https://www.youtube.com/watch?v=vYZbwJJk_hc - **Дата:** 16.11.2025 - **Длительность:** 7:09 - **Просмотры:** 55,773 ## Описание ❤️ Check out Lambda here and sign up for their GPU Cloud: https://lambda.ai/papers Guide: Rent one of their GPUs with over 16GB of VRAM Open a terminal Just get Ollama with this command - https://ollama.com/download/linux Then run ollama run gpt-oss:120b - https://ollama.com/library/gpt-oss:120b 📝 The paper "Fast Physics-Based Modeling of Knots and Ties Using Templates" is available here: https://wanghmin.github.io/publication/guo-2025-fpb/ Sources: https://www.youtube.com/watch?v=2RQcoLV_bVk https://www.youtube.com/watch?v=7d158rQ1R3k https://www.youtube.com/watch?v=qirVdKg3qgs https://www.youtube.com/watch?v=TPokJdN2bkw https://www.youtube.com/watch?v=DRzT3c1jk14 https://www.youtube.com/watch?v=er23-Kt-uHE https://www.youtube.com/watch?v=Odg7acl3nIM https://www.youtube.com/watch?v=jo2ppdJ0Jao 📝 My paper on simulations that look almost like reality is available for free here: https://rdcu.be/cWPfD Or this is the orig. Nature Physics link with clickable citations: https://www.nature.com/articles/s41567-022-01788-5 🙏 We would like to thank our generous Patreon supporters who make Two Minute Papers possible: Benji Rabhan, B Shang, Christian Ahlin, Gordon Child, Juan Benet, Michael Tedder, Owen Skarpness, Richard Sundvall, Steef, Taras Bobrovytsky, Tybie Fitzhugh, Ueli Gallizzi If you wish to appear here or pick up other perks, click here: https://www.patreon.com/TwoMinutePapers My research: https://cg.tuwien.ac.at/~zsolnai/ X/Twitter: https://twitter.com/twominutepapers Thumbnail design: Felícia Zsolnai-Fehér - http://felicia.hu ## Содержание ### [0:00](https://www.youtube.com/watch?v=vYZbwJJk_hc) Segment 1 (00:00 - 05:00) Video games have a clothing problem, and not a small one. You see, in many games, you can choose the clothing for your own character. But often, a piece of clothing simply does not sit well on the character. And I don't mean it's stylistically bad. No, no. I mean, when the game does not simulate it correctly. And sometimes it gets worse. It's not even visible. Ooh, that is really fun. Especially in games where they try to sell you this clothing. So, you paid for nothing. Glorious. And it doesn't matter how much detail you have modeled it with. If it does not sit well on the game characters, no one's going to believe it. So, how can we make them sit well? Well, this research work promises that. So, yes, finally. But I am a little skeptical because knots and ties are notoriously difficult to simulate properly. One designer who does it by hand says that knots fall apart so many times he was close to despair. The reason is that things intersect into each other like crazy and it's nearly impossible to resolve. So, this new research work promises us a physics-based way to solve this not anti problem. It has an editor where we can add a new knot. Okay, I already have a problem here. And the problem is that it looks terrible. Of course, it does. We just drew something and this is not how things look in reality. But then check this out. Wo! It runs a little physics simulation and suddenly it pops into life. Okay, are you ready? Now hold on to your papers, fellow scholars, and it's simulation time. Yo, this looks beyond amazing. Wow. And it all really behaves as it should in reality. So here you can basically just design roughly what the scarf should look like. And then once again, let's go. Yes, it pops into place. This is glorious. And if you look at the other examples of the simulated characters, holy mother of papers. Look at that. This is beautiful beyond words. How is this even possible? But it gets better. It gives you artistic control over the results. And it works on a wide variety of clothing. high resolution models, crazy knots and ties built from hundreds of thousands of vertices, and not one issue or artifact appears. Wow. And get this, this is a completely Michelin star handcrafted technique. No AI was used whatsoever here. Only human ingenuity. Yum, yum. The paper has lots of jargon and mathematics. I mean, of course it does. It was written for other experts and I think I found a way to explain it so anyone would understand what this does. How the magic happens. It's not easy because it is talking about BVH for broadface calling. That's a bunch of words. Okay, so let's do this. Dear fellow scholars, this is two minute papers with Dr. Carola. Dr. Carol. Now, instead of simulating every thread and every fold like a crazy person, instead think of it like giving the computer a straw. It's defined as a basier curve which enables you to bend and twist this straw easily and smoothly. Now, the algorithm takes a look at this straw and it makes it thicker or thinner to avoid weird intersections and weird geometries. Then the puppy thing, the physics simulation shakes it into place so it gets a natural shape. Man, this is brilliant. But we are not done yet. Not even close. It then applies continuous collision detection to make sure things don't intersect into each other. And surprisingly, it doesn't do it looking frame by frame like a human would. Nope. No, sir. It predicts collisions before they happen and corrects them instantly. But we know that this is bloody expensive. So how and now that is what the BVH helps you with. It's called a bounding volume hierarchy. And when there is lots going on that might be colliding, it helps us narrow down the search. How? Well, it does not intersect a highresolution cloth against another highresolution cloth. That is a nightmare. No, instead it puts these cloths into nice little boxes. I'm not even kidding. And finding where these boxes collide is super easy and fast, too. And then when you find two boxes ### [5:00](https://www.youtube.com/watch?v=vYZbwJJk_hc&t=300s) Segment 2 (05:00 - 07:00) overlapping, you perform precise collision tests, but only there. H genius. This is a godsend, saving a ton of computation time. I'll tell you about how long this takes to run and its limitations in a moment. And once again, I don't see anyone else talking about this paper except this one fellow scholar. I salute you. So, help me save a paper today. Like, subscribe, and leave a really kind comment. Okay, so run time. I thought, you know, you wait a couple minutes for this whole thing to run. Apparently not the case at all. It runs in real time. That is jaw-dropping. What a time to be alive. You can pull up a Lambda GPU instance and run simulations like this on it really fast. Now, even this technique has its limitations. It works with templates for these straws. And while [clears throat] new and unusual styles are possible, it takes some modeling in external tools. The other limitation is pretty surprising. Other techniques typically have trouble when the resolution of the garments is too high. Here, not so much. Here, the problem is when you have the opposite. If the 3D cloth model isn't detailed enough, if it has too few triangles, it might poke through itself. This work still handles it better than most, but it might still happen. So, make sure to give it the good stuff and try Lambda. And I got to say, I can't stop playing with OpenAI's Open GPT model through Lambda GPU cloud. And as you see, I am doing very useful things with it for science. Yes, this is actual speed. I can't believe that I can have more than 100 billion parameters running super fast here. Many of you fellow scholars are using it and if you don't, make sure to check it out. It costs only a couple dollars per hour. Insanity. You can rent an Nvidia GPU through lambda. ai/papers --- *Источник: https://ekstraktznaniy.ru/video/11999*