Unreal Engine 5.3 - Next Level Tech Is Coming!

Yes, we just talked about Unreal Engine 5.2, a popular game development platform approximately a month ago. And now the new 5.3 version is already here. So what's new here? Can this really be improved so easily? I mean the previous version put up a clinic in simulating pretty much everything we could think about. In their demo we could drive a virtual car and even the tire pressure and compression was simulated. And so much more. And now 5.3 gives us everything that a wise fellow scholar could ask for. Better explosions, ray traced effects, better class simulations and more. Yummy. First now hold on to your papers fellow scholars, because they now support sparse volume textures. Oh my yes. This enables us to create or bring in already existing volumetric simulations of smoke, explosions and more. And it also supports OpenVDB. This is a software library that helps us compress down the immense amount of data that is required for these simulations. Not so long ago we looked at Nvidia's amazing Neural VDB paper, a variant of this that can cut down the memory footprint of such a simulation by get this up to 100x. 100x less memory needed. Wow. And now we can not only create or import these simulations, but we can also render them. That is light transport ray tracing if you will. Something that we talk about all the time here on two minute papers. Here we start out with a noisy image and over time it clears up. Or if we don't have the time, we can use a noise filtering algorithm to fill in the missing details. This is typically for offline applications here and not for real time games, but typically for feature length movies. But don't forget, real time ray tracing is also possible and it might come to Unreal Engine quicker than we think. They also have a new skeletal editor where I really liked that. Now you can also paint these skinning weights right on the body. Ok, but what does that mean? Skinning weights determine the influence of joints and bones on the body and how the body should move and deform around them. This can now be done so much more efficiently directly in Unreal Engine. Class simulations are now based on the XPBD paper Extended Position Based Dynamics, a legendary Nvidia paper from 2016. It finally made it to the industry and into the hands of millions and millions of users. I am so happy to see this kind of tech transfer happen in and it really shows the world that these papers that you hear about here are real. They get out there in the real world and make it better. Note that they also have an incredible AI based class simulator too that can simulate muscle movements too. And these results are truly something else. Wow. They also introduced orthographic rendering to Unreal Engine. This helps us eliminate the distortion that arises from the shape of of real camera lenses leading to a really distinctive look. This is going to be great for not only the architectural visualization projects, but for top down strategy games too. And there is so much more. They added camera rig rails that makes your in game cameras move a little more like a real cinematic camera. And these amazing virtual LED production stages can now run with higher resolution and higher frame rates, wide angle lenses, multi camera shoots at the same time. Also something that I don't hear too many people talking about is support for foveated rendering. This is for those virtual reality headsets that can track where you look and create more detailed geometry around that spot and do less work on the parts you are currently not looking at. Something that we talked about almost 400 episodes ago and now once again it is making its way into the hands of the whole world. Once again. To my joy, the papers coming alive. What a time to be alive. Lighting with volumetric fog is also better. These beautiful multilayer materials are rendered better. So cool. This is going to be a huge improvement for anyone creating video games, animation, movies, virtual worlds and many other creative things. So now let the experiments begin. If you're looking for inexpensive cloud GPUs for AI, Lambda now offers the best prices in the world for GPU cloud compute. No commitments or negotiation required. Just sign up and launch an instance and hold on to your papers because with the Lambda GPU cloud you can now get on demand H100 instances for just $1.99 per hour and yes, $199 and they are one of the first cloud providers to offer publicly available on demand H100 access. Did I mention they also offer persistent storage? So join researchers at organizations like Apple, MIT and Caltech in using Lambda Cloud instances, workstations or servers. Make sure to go to lambdalabs.compapers to sign up for one of their amazing GPU instances today. This was 2 Minute Papers with Dr. Subscribe if you wish to see more.