Adobe tests AI tools to modify photos, audio, and characters

Adobe showcases several AI prototypes for creation at Max. They transform photos into 3D scenes, help build characters, and automate retouching in Photoshop.

Adobe unveiled several generative AI experiments targeting photo, video, and audio at its Max event. These tools, still in prototype stage, allow navigating an image in 3D, shaping a character by interviewing it, and automating retouching in Photoshop. The message is clear: software-assisted creation is stepping up, but without guarantee of immediate availability.

Prototypes designed to simplify very technical tasks

The first tool, Project Screen Test, is used to build AI characters for video. The user describes a basic personality, then refines the result by talking with the character, like in an audition interview. Adobe states you can specify tone, story, flaws, fears, or movement tics before generating an animated scene. This remains an autonomous creative exercise, with no confirmed link to the publisher’s existing video tools [to be verified].

Another prototype, Project Sight Span, transforms a flat image into a navigable 3D environment. Photos, paintings, and illustrations are supported, preserving the original visual style. The system generates off-frame areas to allow angle changes, then restores resolution when the user fixes a new view. Adobe presents it as a way to reposition the camera more freely while maintaining image coherence.

The third component, Project Sound Bender, applies the same logic to sound. You just import a video and describe the desired ambiance, then the tool generates synchronized sound effects. A square interface then allows real-time mixing of audio variants. In other words, Adobe tries to make mixing less obscure than some consoles that seem straight out of a spaceship cockpit.

Photoshop serves as a lab for more replicable retouching

With Project Rinse Repeat, Adobe pushes further the automation of repetitive retouching. The tool copies a treatment applied to a subject — skin, shadow, visual adjustment — then adapts it to other people or objects in the same visual. The modifications remain editable because they rely on native Photoshop tools. The logic interests both photo retouching and marketing visuals, where work quickly shifts from artisanal to assembly-line, clean version.

These “sneaks” are not announcements of finished products. Adobe presents them as previews of capabilities that could end up in Creative Cloud or Firefly. Caution is warranted, as the publisher has not said which will move to the next stage. However, the strategic direction leaves little doubt: AI is taking a central place in the group’s growth.

What this says about Adobe’s strategy

According to researcher Gautham Mysore, quoted by The Verge, “Making these interactions more intuitive is also about giving experienced creatives more precise control.” This sentence sums up Adobe’s ambition: make complex tasks more accessible without completely removing control from professionals. The prototypes therefore seek to combine fine control and friction reduction, a delicate balance in creative professions where the tool counts almost as much as the result.

In the longer term, these trials also remind us that resource libraries and certain workflows could be challenged by systems capable of generating, modifying, and adapting content on demand. It remains to be seen what will move from the demo stand to daily use. For now, Adobe is mostly testing the boundary between creative assistant and discreet replacement.

By the numbers

  • 4 prototypes — Screen Test, Sight Span, Sound Bender, and Rinse Repeat, presented at Max.
  • 1 annual event — Adobe Max, the publisher’s experimentation showcase.
  • 2 platforms — Creative Cloud and Firefly, where these tools could arrive.

To read