NVIDIA Introduces Edify: Custom AI Model Training for Developers
Tony Kim Aug 15, 2024 01:58
NVIDIA Edify allows developers to create custom AI models using their own data, significantly enhancing content generation capabilities.
NVIDIA has unveiled Edify, a cutting-edge AI architecture designed to help developers build custom models using their own licensed data, according to NVIDIA Blog. This innovation aims to empower the creative community by providing tools for generating high-quality content across various media types, including images, videos, and 3D assets.
Key Features of NVIDIA Edify
Edify stands out for its multimodal capabilities, allowing it to generate diverse types of content from simple text prompts. The system can create images, videos, 3D models, 360-degree high-dynamic-range imaging (HDRi), and physically based rendering (PBR) materials. One of its most notable features is its training efficiency, enabling the production of high-quality content with fewer images.
Additionally, Edify can fine-tune models to match specific styles or learn particular characters and objects. This flexibility makes it a powerful tool for a wide range of applications, from artistic endeavors to commercial projects.
Applications and Use Cases
An exemplary application of Edify is its integration with Getty Images’ generative AI service. By leveraging NVIDIA AI Foundry, Getty Images has trained Edify on its licensed content, ensuring that no copyrighted characters or products are included. This service allows users to generate and modify images while maintaining commercial safety. Contributors to the dataset also benefit from a new revenue stream, as they receive a share of the profits.
Edify’s capabilities extend beyond image generation. It can also create artist-ready 3D meshes with clean topology and up to 4K PBR materials. These meshes are ideal for prototyping scenes, generating background objects, or serving as a starting point for 3D sculpting. The system’s fast preview mode can produce results in as few as 10 seconds, which can then be refined into full 3D models.
Advanced Features for Image Editing
Edify Image offers advanced features for image editing, such as InPaint, which allows users to add or modify content within an image. The Replace feature, a more stringent version of InPaint, can change details like clothing. OutPaint can expand an image to match different aspect ratios, and the Segment feature simplifies object masking with text prompts.
The system also supports advanced prompt adherence and camera controls, enabling users to specify focal length or depth of field. ControlNets like Sketch and Depth guide the generation process, allowing for highly controllable and customizable outputs.
360-Degree HDRi and Multimodal Capabilities
Edify 360 HDRi generates environment maps of natural landscapes that can be used for scene lighting, reflections, and backgrounds. The model can create up to 16K HDRi images from text or image prompts, saving users hours of searching for suitable backplates.
One of Edify’s unique strengths is its multimodal capability, which enables advanced workflows combining different asset types. For example, users can prototype a full scene in minutes with a simple text prompt, as demonstrated in NVIDIA’s Research SIGGRAPH demo. By generating the scene in 3D, artists can frame their desired shot and then use Edify Image to turn the prototype into a photorealistic image.
Generative AI by Getty Images
Getty Images, a leading provider of creative visuals, has utilized NVIDIA AI Foundry to train an Edify Image model for its generative AI service. Available through Generative AI by Getty Images for enterprises and Generative AI by iStock for small businesses, the service allows users to generate and modify images using models powered by Edify.
The latest update to Edify Image has improved generation speed and prompt adherence, introducing new camera controls. Users can now edit and modify iStock’s library of visuals to rapidly iterate and perfect content. These capabilities will soon be available on Getty Images’ platform as well.
For more information, visit the NVIDIA Blog.
Image source: Shutterstock