NVIDIA Researchers Create an Immersive Desert World Using Real-Time Gen AI. The demonstration highlighted the NVIDIA Omniverse platform’s and models’ helpful 3D world-building capabilities when driven by Edify NVIDIA.
SIGGRAPH Real-Time Live event, NVIDIA researchers demonstrated how to use Edify NVIDIA, a multimodal architecture for visual generative AI, to quickly create a detailed 3D desert scene.
In one of the most esteemed sessions of the prominent graphics conference, NVIDIA researchers demonstrated how, in just five minutes, they could create and modify a desert landscape from scratch with the help of an AI agent. The live demonstration demonstrated how generative AI may help artists by speeding up ideation and producing unique secondary elements that would not have needed to be sourced from a repository.
These AI technologies will enable 3D artists to be more productive and creative by radically cutting down on ideation time. They will provide them with the means to explore concepts more quickly and streamline portions of their processes. Instead of wasting hours searching for or generating the 360 HDRi landscapes or backdrop assets the scene requires, they could produce them in a matter of minutes.
Edify NVIDIA
Three Minutes to Go from Concept to 3D Scene
A complete 3D scene takes a lot of work and time to complete. In order to build a rich scene, artists must provide their hero asset with a lot of background objects. After that, they must choose an acceptable background and an environment map to illuminate the scene. They’ve frequently had to choose between producing results quickly and engaging in creative exploration due to time constraints.
Creative teams may accomplish both objectives with the help of AI agents: they can swiftly bring thoughts to life and keep iterating until they have the desired look.
In the Real-Time Live demonstration, the researchers employed an AI agent to direct an Edify NVIDIA-powered model to create previews for hundreds of 3D objects, such as rocks, cactus, and a bull’s skull, in a matter of seconds.
Free 3D Objects
NVIDIA Edify
NVIDIA Edify Quickens the Creation of Environments
Using AI-powered scene generation tools, Edify NVIDIA models can speed up the development of background environments and objects, allowing artists to concentrate on hero assets.
Real Live Time
Two Edify NVIDIA models were displayed in the Real-Time Live demo:
Edify 3D uses text or image prompts to create editable 3D meshes. In a matter of seconds, the model can produce previews that include spinning animations for every object, enabling developers to quickly prototype their ideas before settling on a final version.
Edify 360 HDRi creates up to 16K high-dynamic range photographs (HDRi) of natural landscapes using text or picture prompts. These HDRi photos can be utilised as backgrounds or to illuminate situations.
The researchers also demonstrated USD Layout, an AI model that creates scene layouts using OpenUSD, a platform for 3D workflows, and an AI agent driven by a sizable language model during the demo.
Using Edify NVIDIA-powered technologies, two top creative content firms are providing designers and artists with new ways to increase productivity through generative AI, as NVIDIA announced at SIGGRAPH.
Shutterstock’s Generative 3D service, which enables designers to swiftly prototype and build 3D products using text or image prompts, has debuted in commercial testing. Its Edify NVIDIA-based 360 HDRi generator was also granted early access.
Getty Images
The most recent Edify NVIDIA version was added to Getty Images’ Generative AI by Getty Images service. With twice the speed, users may now produce photos with better output quality, timely adherence, sophisticated controls, and fine-tuning.
3D object Genoerator
Using NVIDIA Omniverse’s Universal Scene Description
USD is a common format used to describe and compose 3D environments, and it is used to structure the 3D objects, environment maps, and layouts created using Edify models. Artists may now import Edify NVIDIA-powered compositions into Omniverse USD Composer right away thanks to this compatibility.
They can further alter the scene in Composer by utilising well-known digital content production tools, such as shifting objects’ positions, altering their looks, or tweaking the lighting.
One of the most anticipated events of SIGGRAPH is Real-Time Live, which offers approximately twelve real-time applications, such as virtual reality, generative AI, and live performance capture technologies. See the replay in the section below.
They then gave the agent instructions on how to use other models to generate possible backgrounds and a plan for the placement of the objects in the picture. This demonstrated how the agent could swiftly switch out the pebbles for gold nuggets in order to adjust to last-minute changes in the creative direction.
They gave the agent a design plan and instructed them to produce high-quality assets and use the virtual world-building program NVIDIA Omniverse USD Composer to render the scenario as a photorealistic image.
3D Objective
Shutterstock Launches Generative 3D, and Getty Images Enhances Their Offering Driven by NVIDIA
With services powered by Edify NVIDIA, you can create 3D assets, light virtual environments, or make photos in half the time. Thanks to generative AI educated on licensed data, designers and artists may now increase their output in new and more efficient ways.
Leading creative content marketplace Shutterstock has put its Generative 3D service into commercial beta. With merely text or image prompts, it enables developers to quickly prototype 3D items and build 360 HDRi backgrounds that illuminate situations.
Leading provider of visual material and marketplace Getty Images has enhanced its Generative AI by Getty Images offering to produce photos twice as quickly, enhance output quality, provide sophisticated controls, and facilitate fine-tuning.
The services are developed utilising Edify NVIDIA, a multimodal generative AI architecture, and the company’s visual AI foundry. NVIDIA NIM, a collection of accelerated microservices for AI inference, is then used to optimise and bundle the AI models for optimal performance.
With Edify NVIDIA, service providers can swiftly scale responsible generative models on licensed data and take use of NVIDIA DGX Cloud the cloud-first approach to leveraging the full potential of NVIDIA AI.
3D Objective
3D modelling is sped up with generative AI
Shutterstock’s service, which is currently in commercial beta for corporations, enables designers and artists to swiftly produce 3D objects that aid in prototyping or populating virtual worlds. By utilising generative AI, for instance, they may rapidly generate the dishes and cutlery on a dining room table, freeing up their time to concentrate on creating the characters that sit on it.
The 3D assets produced by the service come in a number of widely used file formats and are prepared for editing with digital content creation tools. Artists have a sophisticated foundation from which to create their own style because to their simple geometry and arrangement.
In as little as 10 seconds, the AI model first provides a preview of a single asset. The preview can be developed into a better 3D asset using physically based rendering elements like concrete, wood, or leather if customers find it appealing.
This year’s SIGGRAPH computer graphics conference will test the speed at which designers can realise their concepts.
A Blender method that enables artists to create items directly in their 3D environment will be demonstrated by Shutterstock. At SIGGRAPH, HP will display 3D prints and real prototypes of the kinds of assets that guests can create on the show floor with Generative 3D at the Shutterstock booth.
Additionally, Edify NVIDIA 3D generation for virtual production is being used by Shutterstock in collaboration with WPP, a global provider of marketing and communications services, to bring concepts to life.
Best 3D Printed Objects
Would you like to see that stunning new sports vehicle on a winding mountain route, on a tropical beach, or in a desert? Designers are able to change course quickly with generative AI.
Three businesses intend to use the 360 HDRi APIs from Shutterstock straight into their operations. High-end visualisations and 3D material for virtual environments can be created with the 3DEXCITE apps from Dassault Systèmes and WPP, a CGI studio.
