Generative AI has the capability to convert text into images, and this could mean big changes for future photo and video creation.

Spencer Chin, Senior Editor

March 23, 2023

4 Min Read
GettyImages-aiartwork1427969822.jpg
Generative AI is developing the capability to produce photos and artwork based on text inputs. Crystaleyestudio/Getty Images

After decades of using film cameras to take pictures, we have gotten comfortable with using smartphones to take photos and videos and immediately send them to others or post them on social media sites. But the future of image creation may instead lie in using text phrases to guide the creation of still and even moving images.

This is the promise of generative AI, a form of AI that has drawn a lot of publicity in recent months as ChatGPT and other generative AI programs have caught everyone’s attention. Asides from getting instant answers to questions and being able to write entire articles, generative AI programs have the ability to, relatively quickly, produce images of products or people given some inputs, with results that while not exactly artistic can in some cases be a reasonable approximation of a result.

Language is the Basis

The idea of using text to create images may sound far-fetched, but using language models in conjunction with learning and training makes image creation a possibility, according to Bryan Catanzaro, Vice President, Applied Deep Learning Research, Nvidia, during one GTC session.

“All things we do are encoded in language,” said Catanzaro. “There are ways to train a model, given the past words and predicting what the next words will be.”

Related:NVIDIA Omniverse Brings BMW a 30 Percent Boost in Production Planning Efficiency

For image generation, Catanzaro said the continued development of generative AI will enable the construction of images from text and data extracted from existing images, with the algorithms being able to learn given the data provided.

At this week’s Nvidia GTC Developer Conference, the concept of image creation through textural input came up in several sessions. While no one is yet ditching their costly professional cameras and image manipulation programs, the threat of text-based image creation is getting lots of attention, including from companies whose livelihoods rely on current image creation technology.

Getting Adobe’s Attention

One of those companies is Adobe, creators of Photoshop, Illustrator, and several video creation programs.

“Generative AI is changing outcomes for sure,” said Scott Belsky, Chief Product Officer for Adobe, during an online chat session with Nvidia’s Bryan Catanzaro, during GTC. “Generative AI has the ability to generate something very quickly, for those who don’t require pixel perfection.”

Belsky believes AI could also help customers more involved with the creative process, by automating the mundane work and giving them more time to examine possibilities. “This capability can also raise the bar. We are trying to anchor ourselves with customer issues, identifying which things are the most burdensome.”

Related:Next-Gen GPUs Thrust Nvidia Ahead in World of Simulation, Games

Where Belsky believes AI could have the greatest impact is the ability to create an image without an actual physical product.

“We could create an asset and quickly modify it for a variety of customers,” he said. “It may be easier to render a product with photorealistic textures and materials, as well as conceive new products.”

The specter of being left out of generative AI-aided image creation is being recognized not only by Adobe, but also photo services. At GTC, Nvidia announced that Adobe will build models for next-generation creative workflows. At the same time, photo services Shutterstock and Getty Images along with several other companies, said they would use Nvidia’s AI Foundation Cloud Services to customize models for AI-powered applications.

The companies would leverage Nvidia’s NeMo language service and Picasso image, video, and 3D service to build proprietary, domain-specific, generative AI applications for intelligent chat and customer support, professional content creation, digital simulation and more.

IP Hurdles

Besides the potential disruption to current image creation processes, AI also brings other issues. The most important are related to IP and digital rights.

Given that images and artwork are often protected by use rights with the permission of say, a stock photo service, measures likewise need to be taken with ai-generated photos and artwork.

“The more valuable the data, the more proprietary it can be considered,” said Bryan Catanzaro of Nvidia. “There will be a need for data models to be specialized and protect proprietary and confidential data.”

Spencer Chin is a Senior Editor for Design News covering the electronics beat. He has many years of experience covering developments in components, semiconductors, subsystems, power, and other facets of electronics from both a business/supply-chain and technology perspective. He can be reached at [email protected].

About the Author(s)

Spencer Chin

Senior Editor, Design News

Spencer Chin is a Senior Editor for Design News, covering the electronics beat, which includes semiconductors, components, power, embedded systems, artificial intelligence, augmented and virtual reality, and other related subjects. He is always open to ideas for coverage. Spencer has spent many years covering electronics for brands including Electronic Products, Electronic Buyers News, EE Times, Power Electronics, and electronics360. You can reach him at [email protected] or follow him at @spencerchin.

Sign up for the Design News Daily newsletter.

You May Also Like