The Worst Bug In Games Is Now Gone Forever

What this episode is about

The content discusses a groundbreaking advance in digital geometry that addresses a persistent issue in gaming and film: the clipping of thin objects like cloth and ribbons through one another. By utilizing a new piece of research, users can now simulate millions of collisions without any clipping, significantly improving the fidelity of animations and potentially revolutionizing how developers approach these challenges in both industries.

Key takeaways

• Gaming's clipping issues are so notorious, even speedrunners consider it a feature, not a bug.
• VFX teams spend months fixing what a simulation can solve instantly—welcome to the digital age of 'spaghetti'.
• Forget AI—human ingenuity is the new hero in collision avoidance, proving tech isn’t always about neural networks.
• Instead of just wrapping objects in 'bubble wrap', the new method ensures they dance gracefully around each other.
• Who knew 168 million collisions could happen without a single overlap? Welcome to the magic of advanced geometry.

Who speaks in this episode?

• Dr. Károly Zsolnai-Fehér (Host)

What is discussed, and when

1. 0:00 – 3:00: Two Minute Papers, Dr. Károly Zsolnai-Fehér
2. 6:00 – 9:00: Zozo, Offset Geometric Contact, Dr. Ryoichi Ando, SIGGRAPH Asia
3. 9:00 – 11:42: Two Minute Papers

Notable quotes from "The Worst Bug In Games Is Now Gone Forever"

• But it’s written for the very best experts in the world. Well, I am not among those people, but I’ll try my best to explain it in a moment, but I’ll tell you in advance that it…

  about Two Minute Papers · at 2:38

• Now, wait a second. We are wise Fellow Scholars here, so we remember the Offset Geometric Contact paper, OGC. It did something similar. So how does this relate to that? Okay…

  about Offset Geometric Contact · at 6:09

• And it gives me great joy that this research is a one-author, yes a one-author paper, so who is that incredible scientist? Oh my, look at that. It’s our man, Dr.

  about Dr. Ryoichi Ando · at 7:53

Who and what is mentioned in this episode?

People

• Dr. Károly Zsolnai-Fehér — Person · Host / main speaker · 1 mention · first at 2:58
  He is the narrator and presenter of the Two Minute Papers episode, explaining the featured simulation research and framing why it matters for graphics/VFX and industry.
  But it’s written for the very best experts in the world. Well, I am not among those people, but I’ll try my best to explain it in a moment, but I’ll tell you in advance that it…
• Dr. Ryoichi Ando — Person · Referenced expert / paper author · 1 mention · first at 8:00
  Identified as the sole author of the highlighted research enabling extremely robust collision handling ("cubic barrier" approach) and previously known for adaptive fluid simulation work.

Products & tech

• Two Minute Papers — Product · Broadcast · 2 mentions · first at 2:58
  The podcast itself, hosted by Dr. Károly Zsolnai-Fehér, discussing cutting-edge research in graphics and simulation.

Concepts & topics

• Offset Geometric Contact — Concept · Referenced concept · 2 mentions · first at 6:09
  A previous technique in graphics research that added a tiny offset layer around objects to prevent clipping, now surpassed by the new cubic barrier method.

Places & events

• SIGGRAPH Asia — Event · Referenced event · 1 mention · first at 8:37
  The top graphics research conference where the discussed paper was published.
  Hold the phone. Why is a Japanese fashion e-commerce giant publishing at SIGGRAPH Asia, the top graphics research conference? It makes zero sense, until I realized from their…

Sponsors

• Zozo — Company · Sponsor · 2 mentions · first at 8:28
  A Japanese fashion e-commerce giant that published the research at SIGGRAPH Asia, aiming to automate clothing production.
  Hold the phone. Why is a Japanese fashion e-commerce giant publishing at SIGGRAPH Asia, the top graphics research conference? It makes zero sense, until I realized from their…

Full transcript with timestamps

0:00 I can hardly believe that this is possible now, but it really is. Pinch me. You, see gaming has a big problem. And that problem is that things clip into each other all the time. Sometimes it is so bad, there are gamers who specialize in finding these bugs, clipping through walls and doors, and skipping through whole areas to speedrun the game. Yes, the only people who love clipping more than barbers are speedrunners. Surprisingly, the same problem shows up in movies too - but usually you don’t notice it because VFX artists spend weeks if not months fixing those by hand before the movie is released. Studios throw huge compute farms and armies of technical directors at it just to make sure that Superman’s cape doesn’t pass through his neck halfway through the shot.

0:55 Every cape is one bug away from becoming a scarf. So what is the problem? The problem here is that when the geometry of two thin objects, like cloth, ribbons, or our favorite, noodles, touch each other in the digital world, they often sneak through or get stuck. And now we are going to do the impossible. What if I told you there is a completely freely available piece of research where you can put together an insane amount of these little noodles. And see if they clip into each other. Yes, we are literally solving spaghetti. This one showcases millions of collisions, and my goodness…not a single one of them slipped through. This is absolute insanity that all this is possible. Try to speedrun this one! Ha. Checkmate.

1:50 But that’s nothing. We are also going to mercilessly crush some ribbons. Then, you know what? More twisting! Not just one thing, but you know what? Now hold on to your papers Fellow Scholars, and make it five! This goes up to 168 million collisions. Then, we’ll play with these squishy spheres, torture a poor armadillo, and do so much more. Wow. This is incredible. And you can even try it yourself through the link in the video description, if you are a bit technical. Plus I got you a huge surprise in the end. I found something I never expected.

2:32 So how the heck is all that possible? Well, the research paper tells us exactly, but it’s written for the very best experts in the world. Well, I am not among those people, but I’ll try my best to explain it in a moment, but I’ll tell you in advance that it uses no AI whatsoever, this is all human ingenuity. Let’s dive in. Dear Fellow Scholars, this is Two Minute Papers with Dr. Károly Zsolnai-Fehér. Dr. Carroll. So the input here is just the geometry - the mesh of the objects, and the output is a simulation where no matter how many millions of contacts happen, not a single one passes through the other. It’s like trying to juggle thousands of glass marbles and making sure none of them ever overlap or fuse into each other.

3:25 But this one can do even crazier things. Like twisting this piece of cloth into oblivion. It is still going…and still going. Crazy. And it is so accurate, even if we look more closely, I can’t believe it. It is still stable! Wow. And not only that, but look at how beautiful that is. You cannot see this kind of thing anywhere else. So at this point, I’m dying to know - how does this technique do this incredible magic? Well, two things. One, instead of using the old "logarithmic barrier" trick, which panics and freezes when objects get very close, this new method uses a "cubic barrier." So what does that mean? Well, instead of slamming the brakes when two surfaces are about to touch - like the old logarithmic method - this new system eases into the collision with a smoother force curve. Imagine our spaghetti bowl again: instead of waiting until two noodles are fused and then yanking them apart, instead it

4:33 creates a little elastic bubble between them that dynamically adjusts, so they slide past each other gracefully. It’s like adding soap to the spaghetti water - suddenly, everything untangles itself. And two, the paper says that it uses a 3x3 Jacobi block preconditioned Conjugate Gradient method. That sounds crazy, so what is that? Let me try to explain. This is a way to efficiently solve the massive mathematical equations that describe all these forces and movements. Think of it like coordinating a huge, multi-person dance routine where everyone has to avoid bumping into each other. A regular method would try to tell every single dancer exactly where to move all at once, which is chaos. The "3x3 Jacobi block" part means we break the dance floor into small groups of 3x3 dancers and tell each group what to do independently.

5:38 Then, the preconditioned Conjugate Gradient is like a smart coach who quickly refines these instructions for all the groups, making sure everyone moves smoothly and no one crashes, without having to restart from scratch every time. It’s a specialized, iterative way to get everyone moving in harmony without having to calculate every single minute detail at once. A really technical paper explained in simple words anyone can understand. Now, wait a second. We are wise Fellow Scholars here, so we remember the Offset Geometric Contact paper, OGC. It did something similar. So how does this relate to that? Okay, you are going to love this one. This OGC technique was a big step forward because it added a tiny offset layer around objects, like wrapping everything in bubble wrap, so they don’t pass through each other. Of course, that trick worked well for many cases - but when the gaps get extremely tiny,

6:45 like many of these tough cases, OGC still struggles, especially with thin shells or when millions of contacts happen at once. Now get this, this new cubic barrier method doesn’t just wrap things - it actively adjusts its stiffness using the material’s own elasticity, so it can keep even microscopic gaps open. In other words, OGC puts a safety cushion between objects, but this new method is like using a memory foam cushion - it adapts on the fly, reshaping itself to keep even the tiniest gaps open, so the whole system doesn’t tear itself apart. Genius.

7:32 And I absolutely love that all this runs on not a huge data center somewhere, it is just one graphics card. The thing that I rent one of at Lambda. Now of course, you need to be patient, how patient? Well, not seconds per frame, but minutes per frame patient. And it gives me great joy that this research is a one-author, yes a one-author paper, so who is that incredible scientist? Oh my, look at that. It’s our man, Dr. Ryoichi Ando, who created these super cool adaptive fluid simulations before where computation was concentrated around parts where the action really happens.

8:18 This work is a treasure, and I am happy to be alive on this planet and to be able to enjoy it. So good. What a time to be alive! Now, here is the unexpected thing. It is published in the colors of Zozo. Wait wait wait. Hold the phone. Why is a Japanese fashion e-commerce giant publishing at SIGGRAPH Asia, the top graphics research conference? It makes zero sense, until I realized from their press release, aha! Because they want to automate clothing production. Imagine skipping months of sewing prototypes and instead simulating how fabrics drape, wrinkle, and collide - all running on the computer. Of course, to do that, your simulation must never let two layers of cloth clip into each other. This cubic barrier research makes that possible, at scale. So it’s not just academic fun, this is the papers coming alive in the world around us. It’s a stepping stone toward less wasted fabric, faster fashion design, and automated digital tailoring.

9:35 They also have a cloth fitting example. Finally, a way to see if those jeans fit without having to do a small workout in the dressing room! More time to read papers! Okay, limitations? Well, it’s accurate, but slow. How slow? Well, think of it like watching paint dry, except the paint is a million silky noodles colliding. It’s like an orchestra that plays a beautiful symphony, but there is one problem. It is playing at one note per minute. And you know what the craziest thing is? I think this paper is going to change a lot of things in the industry, and absolutely nobody talks about it. This is deep technical wizardry available for free for all of us, and nobody is talking about it. Insanity. This paper is like endangered species, if we don’t talk about them here, I am worried that no one else will. So if you wish to help us save these endangered species, save a paper today, please like,

10:40 subscribe, hit the bell icon, and leave a really kind comment. We are currently at episode 996, we are close to 1,000 paper videos now so we’re not going anywhere. Join the paper club. Noodles, ribbons, spheres, and capers, no collisions with 168 million neighbors. Subscribe to Two Minute Papers.

Questions about this episode

Who speaks in "The Worst Bug In Games Is Now Gone Forever"?

Dr. Károly Zsolnai-Fehér (Host).

What is "The Worst Bug In Games Is Now Gone Forever" about?

The content discusses a groundbreaking advance in digital geometry that addresses a persistent issue in gaming and film: the clipping of thin objects like cloth and ribbons through one another. By utilizing a new piece of research, users can now simulate millions of collisions without any clipping, significantly improving the fidelity of animations and potentially revolutionizing how developers…

What is mentioned in this Two Minute Papers episode?

6 people, works and ideas are mentioned, including Two Minute Papers, Dr. Károly Zsolnai-Fehér, Dr. Ryoichi Ando, SIGGRAPH Asia, Offset Geometric Contact. Each one on this page links to its own page with every episode that mentions it.

How long is "The Worst Bug In Games Is Now Gone Forever"?

The episode runs 12 min; the transcript is about 1,542 words in 12 timestamped sections.

More from Two Minute Papers

• All analyzed episodes
• People mentioned on Two Minute Papers