We got an early look at Adobe’s research projects: How AI is making photo editing faster than ever

Every year in Adobe Maxthe software giant’s annual creative conference, showcases a number of research projects in development. The stealth, as the event is known, is a great way to peek under Adobe’s wide curtain and understand where the tech giant might be heading in the future. At this year’s event in Los Angeles, I got an exclusive early first look at three of my photography projects.

This year’s collection of photography-focused presentations tackles a task that can be handled with manual editing. But instead of spending hours relighting images, removing distractions, or moving 2D objects, these new tools make adjustments in just one or two clicks. These are impressive feats, and they’re made possible by Adobe’s Firefly AI models running in the background.

Trace Erase is one such project. The new tool is based on Adobe’s generative removal technology, which acts as an AI-powered eraser. In Photoshop, you can use generative removal to clean up images with photo bombs, distracting wires, cables, and other errant objects that are obstructing your shot. But it can take several steps to completely remove an object from an image, as Lingzhi Zhang, an Adobe research scientist, showed me in a demo. There are often other signs that something has been removed, such as a reflection in a window. Trace Erase takes care of all these follow-up edits, and as the name suggests, it removes every trace of the object’s presence.

It’s more than just removing reflections and smoothing out what’s left. AI has a broad knowledge base, so if you ask it to remove a jet ski, it will remove the entire object and the waves it created in its wake. If Generative Removal is an AI eraser, Trace Erase is an AI eraser.

There is another graphic hack called Light Touch. With this tool, you can completely change how the image is lit. You can literally use it to change the lighting — Adobe research scientist Zhixin Shu showed me how it can be used on an image of a lamp to turn on its light and adjust how bright it is, automatically aligning the shadows with the environment. You can also diffuse the light to adjust the pop of shadows.

Spatial lighting, the third method in the project, allows you to place the light source within the image. You can make light appear to be coming from any point in the image, including from within existing objects, such as a lit, carved pumpkin. It is an incredibly powerful tool that can transform any photo simply by changing the lighting condition – something that is very difficult to do manually without having to reshoot.

Turn Style is Sneaks’ third photography project. Its name is very similar to the turnstiles you see at events, which makes sense for an AI-powered tool that takes 2D objects and turns them into 3D. Adobe research scientist Zhiqin Chen showed me how the 3D View button can turn a 2D image into a fully rendered 3D object, which you can rotate a full 360 degrees. It keeps the character consistent, so you don’t have to worry about having too many AI hallucinations. While AI helps transform it into a 3D object, you can still control its placement and visual angle.

Some of Sneaks’ previous research projects have eventually ended up in the hands of users, like last year’s project Photoshop AI formatting tooland is now generally available. But there is absolutely no guarantee that this is the case for all of them. What this year’s projects show is that Adobe is focusing its AI efforts for photography and image editing on more practical use cases.

“In Photoshop, it’s not about just clicking a button to view art,” said Steven Nielsen, senior director of product management at Photoshop. “It’s about giving you all the tools to make it easier to achieve what you want to achieve.”

For more, check out AI assistants are coming to Photoshop and Express and Firefly’s first AI music model.