Відео

I dont know what this is for but, Awesome!

ъ.Ъ

So this would be used like baking a texture? But for a whole shader? This seems super interesting. Does it depend on a static scene or is it trained "for any lighting"?

I'm really impressed, i want to ask if it is possible to use now?

We need real ms measurements, hardware, resolution, surface area of the frame that's shading with neural shaders. Yes this is supposedly faster, but you also admitted that the original materials can barely run in real time. Seems like this is going to lead to another manufactured deficiency in graphics or performance.

Reminded me of raymarching

Where is the notebook in the description?

Big step towards generating an object with enough detail to 3d print the objects.. AI modeling leading to functional 3d objects with enough detail to 3d print is a huge goal...

16K textures sounds good. Gaming GPUs will need a LOT MORE VRAM tho.

Nvidias bad vram policy will limit the usage.

we want real machine learning, not these dumbass chatbots LMAO.

Praise Nvidia!

Didn't they write paper about it like... 9 months ago? I am 100% sure I saw all of this and read at least part of that paper.

Nice.

AMD will try to copy it in 3-4 years :P

is there any way we can combine this with blender in future?

I hope the motion capture from video comes fast

Nice, thanks 😁

Where does NVIDIA find these people?

I wonder how this would respond to non-euclidean renderization (although it is very possible that's not how it works and i misunderstood the function of this since most of these concepts are too advanced for me)

No one will feel heavy textures and geometry in fast paced action games, but in, for example, architecture or industrial design they would be much more useful.

The demo is cool, this will be a great speedup for artists to clean up and finalize/customize. And for designers to quickly iterate for scene looks.

Rare Nvidia video with comment section enabled

Well i cant wait for this shit to come out, amd is just gonna be sidelined lmfao,😂

was unreal engine software used??

3ms to render the hd model at the end? No offense but that's terrible considering you need hundreds of models for an average game scene, and all that needs to fit under 16ms for 60fps

I can stil distinguish the full data VDB vs Neural one. But this is really an impressive job compared to the other compression methods. Shoutout to Nvidia!

😮 wich software is this ????

Insane

What in the fark!!! 😱 I wish I could work in NVIDIA, even if it's a water boy job or janitor, that too, without any pay just to witness the magical work behind the curtains. 😢

I love how Nvidia is not waiting for the rate limiting step which is the consoles in innovative gaming development further every generation!

i swear nvidia is the only company doing any innovation. other software “developers” are so dim that nvidia has to pick up the entire computation load. At what point do people realize the nvidia processor is the main, central, processor not the other guys. I would buy an nvidia CPU before seeing the specs if one was announced.

amazing work

Is it available for Android now?

Can you guys please release FVDB as open access? Every time Nvidia unveils something new, it's always closed and selective early access. AI Workbench, Neural VDB, MONAI Cloud, and more. For me, my application status is always stuck at 'We have received your application and we will review it shortly.' It never changes. Would be awesome if everyone could get access! Thanks!

pretty sure it's using voxels, which is why nothing is said about topology. makes sense, as most diffusion models work pixel by pixel.

Noice

Wow, very impressive! ^-^

This visual generative combined with the possible new 3D Vision AI technology it would be amazing…

I hope this available for Apple SceneKit one day! 🙏

it does not show how to add the client to the computer. the most important thing!!!

C++_CUDA has good documentation. Many thanks for the ability to use in functions __device__ void foo(){A* obj=new A; }; dynamic memory allocation in this way is very convenient. If NVIDIA engineers gave the ability to access obj from __host__ it would be super

Ho Lee Fuk

Wait a minute! So, if I prompt anything, I'll get a well-retopologized 3D model in a second, just like you demonstrated with the skull? This is game-changing! But do we have a limited set of prompts based on your 3D library, or can we prompt anything we want and get it in 3D, well retopologized?

🥲

Unreal

YIKES

Это нам, игрокам в копутеры, надо-надо. Работайте, братья💪🏻💪🏻💪🏻

YOU are a super TV presenter girl from Nvidia, Thank you very much for showing such an extensive educational video, you are beautiful, you did a good job showing such important news about the physical indicators of new technologies for life introduced by Nvidia into the technology sector of different countries on the planet, Nvidia company is super number one all over the planet in innovations and software code in microelectronics