Adobe Max sneaks: photos become explorable 3D scenes and character design becomes a conversation

Adobe has used its annual Max conference to unveil a new batch of experimental AI "sneaks," including a tool that turns flat images into explorable 3D environments and another that lets creators develop AI characters by talking to them. According to The Verge, which published the details on November 11, the four projects continue a Max tradition of showing research that may or may not become shipping products.

Project Sight Span turns flat images into 3D spaces

Project Sight Span takes a two-dimensional image — a photograph, a painting or an illustration — and reconstructs it as a 3D environment that users can move through and reshoot from new viewpoints. The system generates whatever was never in the original frame, such as the area behind the camera, and The Verge reports those invented areas read as natural extensions of the scene. Navigation looks rough, with the publication describing the moving view as pixelated and undefined, but image quality returns once a final angle is locked in. Sight Span also preserves artistic style, and details like tattoos and facial features remain recognizable even after repositioning. The Verge compares the result to a far more aggressive take on Apple's Spatial Reframing, while noting that it stretches what most people would still call a photo.

Project Screen Test builds characters through conversation

Project Screen Test addresses a weakness of AI video generation: characters produced this way tend to lack the deliberate personality that traditional animators build in. Users begin with a few base traits — The Verge's example is "anxious, shy, awkward" — and assign a voice via a preset, a text description or a sample audio clip. The persona is then expanded through a video-call-style conversation covering backstory, temperament, goals, flaws, fears and movement quirks. Once finished, the character can be placed into a described environment or scenario to produce animated screen tests. The Verge found the output had a more consistent and identifiable personality than typical AI-animated characters, though the tool is currently a standalone exercise and it is unclear whether it will feed into Adobe's existing AI video products.

Sound Bender and Rinse Repeat

Two further sneaks target audio and retouching. Project Sound Bender generates sound effects synced to an uploaded video clip based on a text description — "bubbling sci-fi cauldron noises" is the example given — and lays them out on a square canvas where each corner holds a different generated sound, such as gurgling, boiling rocks or crackling electric fire. Dragging a cursor around the canvas adjusts which sounds dominate, and each corner can be swapped for a different effect.

Project Rinse Repeat, built around Photoshop, copies edits from one subject and adapts them to others in the same image. Retouching applied painstakingly to one person in a group photo can be propagated to everyone else needing similar work, and the same mechanism can add shadows to differently shaped products in marketing graphics. Because the tool recreates transferred edits with native Photoshop features, they remain fully editable.

Sneaks, not launches

Adobe frames the projects as previews of capabilities that might eventually reach its Creative Cloud apps or the Firefly platform, with no guarantee any will ship. The Verge notes that nearly all of them rest on Adobe's generative AI work, and that with AI now central to the company's growth strategy, at least some are likely to graduate into official features. A shared thread across the four tools is making complex photo editing, filmmaking and audio workflows feel more interactive and approachable, sometimes almost toylike, while still enabling practical changes. Adobe research scientist Gautham Mysore told The Verge that the intuitive interfaces are also meant to give experienced creatives more precise control, pointing to Sound Bender's sonic exploration and Sight Span's camera positioning as places where professionals can apply their judgment and refine results directly.

Why it matters

These previews indicate where Adobe believes creative software is heading: away from assembling discrete assets and toward directing generated environments, characters and sound. If features like Sight Span ship, work that currently calls for 3D artists, location scouting or stock libraries could collapse into a single image plus a prompt — The Verge explicitly calls the sneaks an omen that AI may eventually replace tools and stock libraries creative professionals rely on. For practitioners, the pitch is that conversational and visual controls will translate into finer control rather than less of it. For the broader market, it is another signal that Adobe is placing generative AI at the center of its entire product line, one experiment at a time.