Adobe Firefly Update Promises AI-Generated Images Without Any Weirdness

The latest edition of Adobe's Firefly generative AI model is out, promising better images and prompt analysis, along with the ability to add new details and work faster.

According to Adobe, Firefly has generated over 7 billion images since its debut in March 2023. It has also helped the nearly 40-year-old image-editing software Photoshop realize a 30% year-over-year increase in users, said Zeke Koch, vice president of generative AI product management.

Now, as competitors like Google and OpenAI push into Adobe's visual turf with gen AI tools like Gemini and Sora, Adobe wants to remind users of its prowess in visual content.

The Firefly Image 3 Foundation Model offers "massive advancements to quality and control," Adobe says.

That includes more photorealistic images thanks to better lighting, positioning and attention to detail -- particularly when it comes to rendering people, including emotions.

If you look closely at AI-generated images, you'll likely notice certain details like hands or clothing aren't quite right, or are sometimes blatantly weird. This is what Adobe sought to address in the latest model -- in part by offering what Koch referred to as "photographic detail."

That would be an improvement from earlier this year, when CNET's Stephen Shankland found that "Firefly images suffered plenty of the common generative AI problems like weird anatomy and unreal rendering of areas where two subjects come into contact."

(For expert reviews of generative AI tools including Gemini, Claude, ChatGPT and Microsoft Copilot, along with all the latest AI news, tips and explainers, see CNET's new AI Atlas guide.)

Adobe also says that Firefly Image 3 has a better understanding of text prompts and scenes, which means you can input long, complex prompts to get precisely what you want.

"We made a new version of Firefly that's about twice as big inside," Koch said. "And because it's twice as big, it has twice as much of an understanding of the world [and] it generates images that are way more photographically detailed."

Firefly Image 3 is available in beta via Photoshop and firefly.adobe.com. That means you can try it out and provide feedback, but you can't use the tool commercially. A spokesperson said it will be more widely available later this year.

On Tuesday, Adobe also announced the addition of new editing features to Photoshop, which are powered by Firefly Image 3. They include the ability to use reference images to generate new content, as well as to create and replace backgrounds and fine-tune images. Firefly is also coming to publishing software InDesign.

The goal is to help users bring their ideas to life more quickly. As a result, even novice users can find their way around Photoshop and become more advanced users more quickly, said John Metzger, director of product management of gentech AI in Photoshop.

Editors' note: CNET used an AI engine to help create several dozen stories, which are labeled accordingly. The note you're reading is attached to articles that deal substantively with the topic of AI but are created entirely by our expert editors and writers. For more, see our AI policy.