KEY POINTS
- Adobe showed four experimental AI projects, called sneaks, at its Max conference.
- The tools cover AI character design, 3D scenes from flat images, AI sound mixing and automated photo edits.
- Sneaks are previews only, and there is no guarantee they will reach Creative Cloud or Firefly.
Adobe has unveiled a new batch of experimental AI tools, known as sneaks, at its annual Max conference, according to The Verge. The projects can turn images into explorable 3D scenes, help users design AI characters by talking to them, and make some photo edits as simple as copy and paste.
The first, Project Screen Test, is meant to make AI-generated characters easier to build for video. Rather than describing a character only through text prompts, users speak with it in an audition-like setting. They start with basic traits such as "anxious, shy, awkward" and choose a voice, either from a preset, a written description or an uploaded audio sample. They can then flesh out the backstory, goals, flaws, fears and movement quirks in a conversation, and finally generate animated screen tests in a scenario they describe. The Verge says it is a standalone exercise for now, and it is unclear whether it will join Adobe's existing AI video tools.
Project Sight Span converts a flat image, whether a photo, painting or illustration, into a 3D environment that users can move through and reposition. The space generates objects and backgrounds beyond the original frame, even when the viewer turns around. The 3D view looks pixelated and ill-defined while navigating, but quality is restored once an angle is chosen. The Verge reports that artistic styles are preserved and that details such as tattoos and facial features stay recognizable.
Project Sound Bender handles audio. A user uploads a video clip and describes the desired sound in a prompt, for example bubbling sci-fi cauldron noises. The tool generates effects and syncs them to the footage automatically. It also shows a square canvas whose corners each carry a different generated sound, and dragging a cursor across it changes which sounds stand out in real time. Users can also replace any corner with a different effect.
Project Rinse Repeat targets repetitive work in Photoshop. It copies an edit made to one object or person and applies it to others in the same picture, such as adapting blemish retouching across everyone in a group photo or adding shadows to product images with different shapes and sizes. Because the tool rebuilds the changes with native Photoshop tools, the results stay fully editable.
The Verge notes that sneaks are technology previews that may eventually become part of Creative Cloud apps or the Firefly platform, but nothing is guaranteed. Almost all of the projects rely on Adobe's generative AI work, which the outlet describes as central to the company's growth strategy. It also says the common thread is making complex photo, video and audio workflows more intuitive.
Gautham Mysore, a senior principal research scientist at Adobe, told The Verge that intuitive interaction also gives experienced creatives "more precise control." He pointed to Sound Bender's sonic exploration and Sight Span's camera positioning as ways professionals can apply their judgment and refine results more directly.
