Shutterstock's 3D generative capabilities are open in early access, Getty Images is rolling out custom fine-tuning capabilities for enterprises, Adobe is bringing 3D generative AI to creators on Firefly and Creative Cloud, and Be.Live, Bria, and Cuebric choose NVIDIA Picasso AI Foundry for inference.
NVIDIA Edify, a multimodal architecture for vision generative AI, is entering a new dimension.
Edify introduces new features for developers and visual content providers, such as 3D asset generation, that enable greater creative freedom through AI image generation.
Multimedia content and data provider Shutterstock's application programming interfaces (APIs) built on the Edify architecture will be available in Early Access. The API allows creators to quickly generate 3D objects for use in virtual sets using prompt text or images.
Getty Images is a visual content creation company and a marketplace for visual content with generative AI services that are safe to use in the commercial world. The service will add custom fine-tuning capabilities to help enterprise customers generate visuals that align with their brand guidelines and style. Other new features will be added to the service to help customers gain more control over the results of their generated images.
NVIDIA NIM is a set of microservices for inference that was launched at NVIDIA GTC 2024. NVIDIA NIM allows developers to try out pre-trained Edify models from Getty Images and Shutterstock as APIs. In addition, NVIDIA Picasso, an AI foundry service built on the NVIDIA DGX cloud, allows developers to train and deploy their own generative AI models using the Edify architecture.
NVIDIA and Adobe are collaborating to bring new 3D generative AI technology built on Edify to millions of Firefly and Creative Cloud creators.
Live streaming platform Be.Live is using NVIDIA Picasso foundry services to deliver real-time generative AI that automatically creates visual effects to deliver a unique interactive experience for viewers. Bria, a full-featured platform purpose-built for businesses developing responsible visual generative AI, has now adopted Picasso to run inference. Cuebric Creative Studio is developing generative AI applications powered by Picasso to create immersive virtual environments for filmmaking and content creation.
Accelerate 3D creation: The Shutterstock 3D AI generator is now available in Early Access
Now available in Early Access, Shutterstock's 3D AI service helps creators generate a variety of virtual objects for use in sets and ideation. This feature can dramatically speed up scene prototyping, giving artists more time to focus on the main characters and objects.
Loading...
Shutterstock 3D generator in action (Video source: Shutterstock)
With these tools, creative professionals can quickly create assets with text prompts or reference images, and export asset files in a variety of popular 3D formats. The Edify 3D service also comes with built-in protections that filter the generated content.
This commercially safe model is trained on Shutterstock's licensed data. Shutterstock has paid hundreds of thousands of artists, and millions more are expected to be paid in the future, given the important role that artists' content intellectual property plays in training generative techniques.
Assets created with the Shutterstock 3D AI generator will be rendered in a tiled composition. (Source: Shutterstock)
HP and Shutterstock announced at GTC that they will collaborate to leverage Edify 3D to provide designers with a wide range of prototyping options to enhance the custom 3D printing experience.
Shutterstock's 3D AI generator helps designers quickly iterate on concepts and create digital assets, while HP automates workflows to transform created digital assets into 3D-printable models. HP 3D printers can then print these models into physical prototypes that help designers inspire their products.
Mattel is using 3D generative AI from Shutterstock to speed up the design ideation process. With AI, toy designers can visualize their ideas for new products with simple text descriptions. By lowering the technical barrier to entry for creating high-fidelity concept designs, designers can explore a wider creative library and iterate faster.
In addition, Shutterstock is developing various Edify-based tools to generate 360 HDRi environments with text or image prompts to illuminate 3D scenes.
Dassault Systèmes is working on 3D content creation through its advanced 3DEXCITE application, integrating Shutterstock's generative 360 HDRi API into its workflows based on NVIDIA Omniverse, a computing platform for developing OpenUSD-based 3D workflows and applications, and CGI studio Katana is advancing the same integration.
Accenture Song, the world's largest technology-enabled creative group, is using computer-aided design (CAD) data in the Omniverse platform to generate high-fidelity images of Defender models for marketing. Using generative AI microservices powered by Edify, Accenture Song was able to build cinematic, interactive 3D environments with conversational prompts to create a 3D scene that provides a fully immersive experience, blending the resulting photorealistic environment with the digital twin of the Defender vehicle.
Take control of the big picture: Turn your visual ideas into reality with Getty Images' custom fine-tuning capabilities
Getty Images will continue to enrich the capabilities of its generative AI services that are safe to use in the commercial world, and provide appropriate intellectual property compensation to users who use the content generated by the service.
At CES in January, Getty Images unveiled an image inpainting API powered by Edify for adding, removing, or replacing objects in images. The API also expands the creative space with expanded imagery. These features are now available on Gettyimages.com and iStock.com.
Beginning in May, Getty Images will also offer a custom fine-tuning service so that the company can adapt the Edify base model to its branding and visual style.
With the service, companies can upload proprietary datasets, view auto-generated labels, submit fine-tuning tasks, and deploy the results to production once they have the results, all self-service and without the use of any code.
Having finer control over image output has been one of the biggest challenges for generative AI. To that end, Getty Images will also introduce a suite of APIs that will be available as part of a custom fine-tuning tool that will give users more control over their image output.
Soon, developers will be able to use the Sketch, Depth, and Segmentation features. The Sketch feature allows users to provide sketches to guide AI-generated images, the Depth feature allows users to copy the composition of a reference image with a depth map, and the Split feature allows users to split portions of an image to add, remove, or improve characters or objects.
Several leading creative agencies and advertisers are already using Getty Images' API services, which include:
- Dentsu Inc. : Japanese advertising agency Dentsu is powering the MAFA: Manga Anime For All app, which creates comic and animation-style content for marketing cases, using Getty Images' generative AI API service. In addition, Dentsu Creative is using NVIDIA Picasso to fine-tune the Getty Images model for Sam's Club, a leading membership-based warehouse retailer.
- McCann: The creative company used generative AI to develop an innovative game for Reckitt's over-the-counter cold remedy, Mucinex, that allowed players to enjoy interacting with the brand's mascot.
- Refik Anadol Studio: Known for its use of generative AI for art, the studio showcased rainforest-inspired artwork at GTC. The installation was created using Getty Images' AI model and fine-tuned Refik's Rainforest Catalog.
- WPP: The marketing and communications services company is working with The Coca-Cola Company to explore how Getty Images' models can be fine-tuned to help build custom visuals that align with the brand's style and guidelines.
Large-scale natural model: Refik Anadol Studio's Living Archive at GTC 2024
Learn more about NVIDIA Picasso and try Edify-powered NIM from Getty Images and Shutterstock at ai.nvidia.com.
The header image on the left shows assets created by Shutterstock's 3D AI generator powered by Edify (Source: Shutterstock). The Edify feature of generating images from sketches is shown on the right side of the header image, provided by NVIDIA.