Adobe previews four experimental AI tools for creative work

By Sophie Lindqvist

The Verge reported on November 11, 2026, that Adobe previewed four experimental AI tools for developing characters, building navigable 3D scenes, mixing sound effects and transferring photo edits. The projects appeared ahead of Adobe’s MAX Sneaks presentation at the company’s annual conference.

The Verge named the projects Project Screen Test, Project Sight Span, Project Sound Bender and Project Rinse Repeat. The publication described them as technology previews that could eventually be developed for Creative Cloud apps or Firefly, and said Adobe has not committed to releasing them.

Across the demonstrations, The Verge said the shared aim was to make complex filmmaking, image editing and audio mixing more interactive without reducing every adjustment to another text prompt. For creative professionals, the stakes are practical: can these direct and revisable controls work inside production files rather than only within polished demonstrations?

What did Adobe preview?

The Verge’s lineup covers four different jobs inside a creative workflow:

• Project Screen Test builds a character from traits and voice before staging an animated screen test.
• Project Sight Span turns a flat image into a navigable 3D environment with content beyond the original frame.
• Project Sound Bender places generated sound effects on a spatial canvas and synchronizes them with uploaded video.
• Project Rinse Repeat copies a retouching or product-shadow treatment across subjects or objects in one photograph.

The tools use different forms of direct manipulation. A conversation shapes a character, moving through a scene changes the viewpoint, moving a cursor changes a sound mix, and copying an edit spreads it across a photograph.

The Verge said this approach tries to keep users in control instead of reducing every change to another text prompt. None of the four projects has an announced release or a confirmed place in Adobe’s product lineup.

Takeaway: The previews put hands-on control first, but they remain experiments rather than announced products.

How does Screen Test create a character?

According to The Verge, Project Screen Test starts with a set of personality traits and a voice. The voice can come from a preset, a written description or an uploaded audio sample.

Users develop the character through an audition-like conversation that establishes its backstory, temperament, goals, fears and movement quirks. A separate Personality tab lets people manually adjust the generated traits.

After that process, The Verge said, the tool can place the character in a described setting or scenario and produce an animated screen test. The Verge found that the results gave characters a more identifiable and consistent personality than many animated AI characters.

Screen Test remains a standalone creative exercise, and The Verge said its relationship with Adobe’s existing AI video tools remains unclear. That leaves open basic questions about how a generated character would move from a personality exercise into a larger production workflow.

Takeaway: A character can gain a stable persona before filming, but Adobe has not fitted Screen Test into a video workflow.

How does Sight Span rebuild a scene in 3D?

The Verge says Project Sight Span accepts a flat photograph, painting or illustration and converts it into a three-dimensional environment that viewers can navigate and reposition.

It generates objects and backgrounds beyond the original frame, including imagery that appears when the viewer turns toward what was behind the original viewpoint. The Verge said this changes the apparent shooting point rather than simply selecting a different crop.

In The Verge’s demonstration, repositioned tattoos and facial features remained recognizable relative to the reference image. The tool also preserved and replicated artistic styles.

Yet the navigable environment appeared pixelated and ill-defined during movement. The Verge said resolution was restored after an angle was selected, suggesting that the system generated surroundings more successfully than it sustained fine detail while the viewer moved.

Takeaway: Sight Span can move beyond a reference image, but its navigation quality still falls short of production-ready visuals.

How does Sound Bender turn prompts into a mixer?

According to The Verge, Project Sound Bender generates sound effects from a text description and synchronizes them with an uploaded video. A text prompt starts the process, but the interface treats the generated effects as elements the user can arrange.

Its square canvas assigns a relevant generated sound to each corner. Moving the cursor around that space changes which sounds dominate in real time, while individual corners can be customized with different effects.

The Verge described the result as a spatial mixer for exploring combinations of sounds rather than choosing just one generated effect. Users can hear how competing sounds interact and adjust the balance without replacing the entire audio track.

Takeaway: Sound becomes a mix that users can shape directly, not a single effect locked behind one prompt.

How does Rinse Repeat reuse an edit?

According to The Verge, Project Rinse Repeat targets repetitive work within a single photograph. The tool applies one person’s blemish retouching to other people in a group and adapts those changes to each subject.

In another demonstration, it added a shadow to one product and reproduced the shadow across products with different shapes and sizes. The same editing process therefore covers both small corrections on faces and repeated commercial design elements.

The Verge said the tool is built around Photoshop and recreates transferred changes with native Photoshop tools, leaving the results fully editable. The source called that distinction important because automation can propagate an editing process without removing the ability to revise individual results.

Takeaway: Rinse Repeat spreads repetitive edits across a photograph while preserving Photoshop’s normal editing control.

What connects the four experiments?

Adobe research scientist Gautham Mysore said the interaction designs also seek to give experienced creatives more precise control. He cited Sound Bender’s sonic exploration and Sight Span’s camera positioning as ways professionals could apply their judgment and refine results directly.

“That connection between creative intent and control is a big part of what we’re exploring,” Mysore said.

Adobe’s official MAX Sneaks session page describes the event as “a first-hand look at work underway in Adobe Labs.” The official schedule put the session at 5:30 p.m. EST on November 11, later than the morning disclosure of the projects.

Takeaway: The common research goal is to connect creative judgment with direct manipulation.

Will these tools reach Creative Cloud?

The Verge framed the practical test as whether precise, editable control will carry into any released versions. Screen Test still lacks a defined place in Adobe’s video workflow, while Sight Span’s demonstration showed degraded visuals during navigation.

The Verge said that leaves both tools’ usefulness in production work unproven, despite their more direct forms of creative control. Adobe has not committed to turning any of the four previews into Creative Cloud apps or Firefly features.

If any preview reaches those products, release timing, production-ready image quality and a clear role in a professional workflow will determine whether the direct controls survive beyond the demo.

Takeaway: Adobe still has to prove that these direct controls can become dependable features rather than polished experiments.