No Spin. Just The Source.
unga.news

Adobe Showed SI That Turns One Flat Photo Into a Place You Can Walk Around In

At its yearly Sneaks show, hosted by Terry Crews, Adobe demoed 10 experiments, from a photo you can step inside to a movie character you can talk to. Two of them are open to try right now.

By @UngaTheGreat ·
Share on XText
Image: Terry Crews, who hosted Adobe’s MAX Sneaks show. Photo: Gage Skidmore, CC BY-SA 2.0, via Wikimedia Commons

Every year at its MAX conference, Adobe saves the strangest stuff for last. It’s called Sneaks, a stage show where its researchers demo half-finished ideas that may or may not ever ship. This year’s show, hosted by actor Terry Crews in Miami, had 10 of them, and a few of them looked like magic tricks.

The one people will remember is called Project Sight Span. You hand it one ordinary flat photo, and it turns that picture into a place you can move around in. Adobe’s own description says it “transforms an image into a navigable environment that creators can step inside and explore, finding new angles and composing shots beyond the original frame.”

In plain terms: you took a photo from one spot, and now you get to stand somewhere else. Swing the camera left. Look past the edge of the frame. Find a better angle after the moment is already over.

Walking around inside a picture

The Verge’s Jess Weatherbed watched the demo and described how it works on screen. While you move around, the 3D view gets blocky and pixelated. When you stop on the angle you want, the detail comes back. She noted that faces and tattoos still looked recognizable after the move, which is the hard part, because the tool has to invent everything the camera never saw.

She called it “Apple’s Spatial Reframing feature on steroids,” and she also raised the fair question hanging over the whole thing. Once a computer has filled in half the scene, the result “feels far less deserving of being called a photo afterward.” That’s worth chewing on. A photographer’s whole job used to be standing in the right spot. Sight Span makes the spot negotiable.

For a family snapshot, that’s a fun trick. For a news photo, it’s the kind of thing that should come with a label.

A character you can call on FaceTime

The weirdest demo of the night, by The Verge’s count, was Project Screen Test. A writer starts with a few traits, like “anxious, shy, awkward,” gives the character a voice from a preset, a typed description or an uploaded sample, and then just talks to it. Weatherbed said the conversation runs “as if you’re talking on FaceTime,” and the tool can spit out animated screen tests of the character.

Adobe pitches it as a way to figure out who a character is before you write a single scene. Ask your villain what she wants. See how your shy kid answers a hard question. It sounds silly until you remember that actors do exactly this in rehearsal, and writers have always talked to their characters in their heads. Now the character talks back.

The rest of the video lineup goes after problems anybody who has edited a home movie will recognize. Project Scene Sweep lets you make an edit once and have it carry across every cut and camera angle in a scene, instead of fixing the same thing shot by shot. Project Live Set lets you change the subject, the setting or the camera while a generated video is still playing. Project On Cue tackles timing, so a flash of light or some other action lands right on a beat that you draw by hand or pull from a song.

Pulling one sound out of the mix

Audio got one big swing, Project Sound Bender. You type the name of a sound, and it lifts that sound out of a finished mix, the kind of job that usually takes a pro and a lot of patience. From there, the sound lands on a square canvas with variations at each corner, and you slide between them to reshape it. The Verge’s example was “bubbling sci-fi cauldron noises.” Podcasters and radio folks were already passing that clip around on 𝕏 within the hour.

A brush that knows where the light is

The photo and design projects are where this gets practical fast. Project Deep Brush reads an image’s shape, color and lighting, so a brushstroke can wrap around the curve of a surface, a color change keeps the natural shadows, and you can literally paint light into a scene. Project Hands On picks out the objects in a picture and puts handles right on them, so you can grab a thing and move it the way you’d move it in real life.

Project Rinse Repeat is the one retouchers will love. Fix a blemish or tweak something in one part of an image, and it repeats that work somewhere else, adjusting for the new spot. Unite.AI described it carrying a retouch across several people in a group photo. Adobe stresses that the edits get rebuilt with regular Photoshop tools, so you can still go back in and change them. That matters, because most of these one-click tools leave you with a flattened result you can’t touch.

Then there’s Project Link Up, which might be the most fun to watch. Drag a car toward an empty spot in an illustrated parking lot, and it resizes, tilts and restyles the car to fit the drawing. What comes out is editable vector art, not a stuck picture.

Two you can try tonight

Here’s what’s new this year. Sneaks used to be look-but-don’t-touch. This time Adobe opened a page called the Adobe Labs Playground, and two of the 10 projects, Deep Brush and Hands On, are live there for anyone to try. There’s also an early look at Project Custom Filters for Photoshop. Adobe is asking people to send what works, and what doesn’t, to labs@adobe.com.

That’s a smart move. Letting regular people kick the tires tells Adobe fast which of these is a real product and which one is just a great stage moment.

Nothing here comes with a promise. DPReview and others pointed out that Adobe hasn’t committed to releasing any of them. Some past Sneaks turned into features people use every day. Plenty never showed up again.

Still, it’s a clear picture of where the photo and video business is headed. The tools are learning what’s actually in an image, the faces, the surfaces, the light and the sounds, and letting you change it by pointing and dragging instead of clicking through menus. Sight Span is the headline grabber. Rinse Repeat and Deep Brush are the ones that could quietly save people hours.

And somewhere in there is a question nobody on stage had to answer: if one photo can become a whole room you walk around in, how long until “I was there” stops meaning much in a picture?

ShareShare on XText

Sources

  1. Adobe — Adobe MAX Sneaks 2026 (Nov. 11, 2026)
  2. The Verge — Adobe’s MAX Sneaks experiments
  3. DPReview — Adobe’s showing off experimental features that may shape the future of Photoshop
  4. Unite.AI — Adobe previews four AI experiments for creative workflows
← Back to unga.news