Adobe Previews Four AI Experiments Aimed at Creative Control

Adobe previewed four experimental AI tools on November 11, 2026, ahead of its MAX Sneaks session — and none of them starts with a text prompt. The Verge reported that Project Screen Test, Project Sight Span, Project Sound Bender and Project Rinse Repeat are technology previews that could eventually reach Creative Cloud apps or Firefly. Adobe has not committed to releasing them.

The shared idea across the demonstrations: make complicated filmmaking, image-editing and audio-mixing work more interactive without reducing every adjustment to another text prompt.

What does Project Screen Test do?

Screen Test builds a character from a set of personality traits and a voice — supplied by a preset, a written description or an uploaded audio sample. Users then develop the character through an audition-like conversation, establishing its backstory, temperament, goals, fears and movement quirks. A separate Personality tab allows manual adjustment of generated traits. Once developed, the tool can place the character in a described setting and produce an animated screen test.

The Verge found the results gave characters a more identifiable and consistent personality than many animated AI characters. The caveat: Screen Test is currently a standalone creative exercise, and its relationship to Adobe's existing AI video tools remains unclear.

How does Sight Span turn photos into 3D scenes?

Project Sight Span converts a flat photograph, painting or illustration 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 approach 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, and Sight Span could preserve and replicate artistic styles. But the navigable environment appeared pixelated and ill-defined; resolution returned only after an angle was selected.

What are Sound Bender and Rinse Repeat?

Project Sound Bender generates sound effects from a text description and synchronizes them with an uploaded video. Its square canvas assigns a relevant generated sound to each corner. Moving the cursor around that space changes which sounds dominate in real time, and individual corners can be customized with different effects. It works as a spatial mixer for exploring combinations rather than choosing just one generated effect.

Project Rinse Repeat tackles repetitive work inside a single photograph:

• It can transfer one person's blemish retouching to other people in a group, adapting the changes to each subject.
• In another demonstration, it added a shadow to one product and reproduced it across products with different shapes and sizes.

The tool is built around Photoshop and recreates transferred changes with native Photoshop tools, leaving the resulting edits fully editable. That distinction matters: automation propagates an editing process while preserving the ability to revise individual results.

Why Adobe says control is the point

Adobe research scientist Gautham Mysore said the interaction designs are intended 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 session was scheduled for 5:30 p.m. EST on November 11, later than the morning disclosure of the projects.

For creative professionals, the practical question is whether precise, editable control will carry into any released versions. Screen Test still lacks a defined place in Adobe's video workflow, and Sight Span's demonstration showed degraded visuals during navigation — leaving both tools unproven for production work despite their more direct forms of creative control.