Adobe previews four AI experiments for photos, video and audio

By Amara Osei

Adobe showcased four experimental AI projects at its annual Max conference: Screen Test, Sight Span, Sound Bender and Rinse Repeat. All four remain prototype “sneaks” (technical demonstrations), with no release dates, pricing or confirmed routes into Creative Cloud or Firefly, Adobe’s generative-AI platform.

For photographers and photo businesses, the distinction between demonstration and product determines near-term workflow decisions. Studios cannot responsibly promise clients access to these tools, so current retouching, video-editing and asset-management systems remain the practical baseline.

What does Screen Test change for video work?

Project Screen Test turns AI-character development into a spoken audition. A user first defines base traits such as anxious, shy and awkward, then assigns a voice through a preset, written description or uploaded reference audio.

The user then interviews the character to expand its backstory, temperament, goals, flaws, fears and movement quirks. After that definition, the user supplies an environment or scenario, and Screen Test generates an animated screen test for video content.

The preview produced a more identifiable and consistent personality than many generic AI-animated characters. That result addresses a central weakness of generative video, which offers less creative control than traditional animation with every character detail deliberately specified.

Adobe currently presents Screen Test as a standalone creative exercise. The company has not explained how, or whether, it will join existing AI video tools, leaving studios without a reliable production handoff.

Can a flat photograph become a 3D set?

Project Sight Span converts a flat image into a navigable 3D environment and lets the user reposition the viewpoint. Adobe generates objects and backgrounds beyond the original frame, extending the scene in a way intended to look natural.

The preliminary 3D output looks pixelated and ill-defined. Adobe restores the image resolution after the user selects a chosen angle, separating the navigation preview from the selected view.

Sight Span accepts any flat image, including photographs, paintings and illustrations. It can preserve or replicate artistic styles, while repositioned tattoos and facial features remain recognizable and close to the original reference.

The demonstration compared Sight Span with Apple’s Spatial Reframing feature “on steroids,” allowing a more aggressive change of shooting point. Yet extensive repositioning can make the result feel less deserving of the label photograph, a limitation relevant to commissioned retouching and campaign delivery.

How do the audio and copy-edit tools work?

Project Sound Bender accepts a video clip and a text request for a sound treatment. It generates effects, synchronizes them to the footage and presents them on a square sound canvas, with each corner assigned a relevant layer.

Users move the cursor to emphasize selected sounds or customize the corners. Example layers include:

• Dense gurgling with popping bubbles
• Rumbling boiling rocks
• Pitched bloops
• Crackling electric fire

The workflow could reduce manual matching of effects to picture in post-production. It remains a preview, however, so agencies cannot yet use it as a dependable delivery standard for client video.

Project Rinse Repeat copies an edit and applies it to other people or objects within the same photograph. Users can replicate blemish removal across a group or add shadows to differently shaped products in marketing graphics.

Because Rinse Repeat recreates each change with native Photoshop tools, the resulting edits remain fully editable. That editability could preserve fast client revisions while reducing repetitive production work.

What changes for creative businesses?

Almost every project builds around Adobe’s generative-AI development, and the company may eventually place some capabilities in Creative Cloud apps or Firefly. Adobe offers no guarantee that any sneak will graduate into a shipping feature.

AI already plays a central role in Adobe’s growth strategy, which raises the likelihood that at least some experiments will reach an official launch. Their wider commercial effect could exceed the demonstrations: if useful enough, AI may eventually replace the tools and stock libraries that creative professionals currently use.

Across the previews, Adobe emphasizes making complicated photo editing, filmmaking and audio mixing more accessible, intuitive and interactive. Some demonstrations resemble toys, but they also show routes to practical changes in active projects.

“Making these interactions more intuitive is also about giving experienced creatives more precise control,” senior Adobe research scientist Gautham Mysore said. “Sound Bender’s sonic exploration capability and Sight Span’s camera positioning are examples of how professionals could apply their judgment and refine a result more directly. That connection between creative intent and control is a big part of what we’re exploring.”

For client work, no prototype yet warrants replacing established software or buying new gear around these capabilities. The next material checkpoint is Adobe identifying launch dates, destination products and commercial terms for whichever experiments graduate from sneaks into shipping features.