Adobe unveils AI features for Photoshop — but not everyone is happy about it

Adobe unveils AI features for Photoshop — but not everyone is happy about it

Adobe has new generative AI features for novice Photoshop users, but there are concerns from creative professionals.

On Tuesday, Adobe announced A beta version of Photoshop with new features for creating and editing images from text prompts. Generative AI capabilities are bolstered by Adobe's Firefly Image 3 model, which was also released today.

New in photo editing software is a generate image tool that creates an image from a text prompt, leaving people dreading the blank page.

In Adobe Photoshop, users can now create entire images from a text prompt.
Credit: Adobe

Generative Fill, a pre-existing tool that fills in the background or expands an image with a wider view, now includes the ability to create an image with a reference image feature. So if you want to create an image with a certain aesthetic, you can upload an image and push the tool to achieve that aesthetic or look.

Diamond studded bee on flower with diamond necklace image

Reference Image lets users reference AI-generated images from existing images.
Credit: Adobe

These tools make it incredibly easy for Photoshop novices (and pros for that matter) to bring their ideas to life. Especially now that Firefly Image 3 is capable of amazing photo-realistic detail and better understanding of gestures. But Adobe has pushed ahead with creative AI features despite controversy over its training data and backlash from creative professionals.

Mashable Light Speed

The problem of training data for AI models is a complex problem that is not going away anytime soon. Companies love it. Google And Open AI have been Trial To train your models on copyrighted works, and this is for text output only. The unveiling of OpenAI's Sora (which is not yet available to the public) and AI video generators just released by Adobe have extended this problem to other sources. While companies are pushing ahead with AI development to catch up with their competitors, many creators as well @Rahll At X, angered by the lack of oversight over technologies that threaten to replace or disrupt their operations.

Adobe has been adamant that Firefly is safe for commercial use, and has touted its model as an ethical alternative to competitors like Midgerny or Stable Diffusion. accused of Train your models on copyrighted works without credit or compensation. A more recent one Bloomberg Reports It turns out that Firefly was trained on AI-generated images taken from Midgerny and other competitors, meaning that Adobe's purportedly “commercially safe” model consists of training data drawn from tasks that many people use. The idea is infringed or stolen.

Adobe's Creative AI models are trained on “licensed content, such as Adobe Stock, along with public domain content,” a spokesperson said. But, according to a Midgerian enthusiast Nick St. PierreAbout 13 percent of Adobe's stock library is AI-generated content that is part of the data used to train Firefly.

That said, the spokesperson said that “training data must comply with our IP guidelines and pass our multi-layered, continuous review and moderation process,” so unlicensed AI-generated There is a moderate amount of material that can slip through the cracks.

X users have pointed out the problem of creating an image based on the work of a photographer or artist, especially when it comes to the reference image tool. “Great. Stolen 'reference photos' from real photographers who went to great lengths and spent a lot to make them.” Posted @notbrodyjenner. “Then pick an artist who has spent his career creating a unique style, rip him off wholesale and make whatever garbage you want. Awesome.”

Adobe compensates partners through stock royalties and a Bonus program which encourages users to contribute to Firefly training.

About the Author

Leave a Reply