Adobe is showing experimental AI projects that let creators shape images, characters, audio, and photo edits through more direct interactions. The demonstrations range from exploring a photograph as a 3D scene to talking with a character before generating an animated screen test.
Character development starts with a conversation
Project Screen Test is designed to help develop AI-generated characters for video. Creators begin by describing personality traits and choosing a voice, writing a description of how it should sound, or uploading sample audio for reference.
They can then talk with the character to develop details such as its backstory, temperament, goals, flaws, fears, and movement quirks. Once that persona is set, a creator can describe a scenario or environment and generate an animated screen test.
The project is currently a standalone creative exercise, and Adobe has not made clear whether it will become part of its existing AI video tools. The demonstration suggests one way to give generated characters more consistent personalities, while also highlighting a limitation: AI video offers less creative control than traditional animation, where each character detail can be intentional.
Images become spaces to explore
Project Sight Span turns a flat image into a navigable 3D environment. Users can move around the scene and change the camera position, while the system generates objects and backgrounds beyond the original frame, including areas that were not visible in the source image.
The tool works with photographs, paintings, and illustrations, and can preserve their artistic styles. During exploration, the scene may look pixelated and indistinct; the resolution improves when the user selects a final viewpoint. The demonstrations also show repositioned facial features and tattoos remaining recognizable.
This gives creators more room to reframe an image or choose a different viewpoint. At the same time, changing a picture’s perspective so substantially may make the result feel less like a conventional photograph.
Sound and edits get interactive controls
Project Sound Bender focuses on audio mixing. A user uploads a video and describes the sound they want, such as “bubbling sci-fi cauldron noises.” The project generates effects, syncs them to the video, and offers a square canvas for adjusting the mix.
Different parts of the canvas can represent different sounds, from gurgling bubbles to rumbling rocks or crackling electric fire. Moving the cursor changes which sounds stand out, and users can customize the sounds assigned to each corner. The interface makes audio exploration feel direct, while leaving room to refine a specific result.
Project Rinse Repeat applies a similar idea to photo editing. After a user edits one person or object, the tool can adapt those edits to other subjects in the same image. A retouch on one person, for instance, can be applied to others who need similar cleanup; edits can also add shadows to products with different shapes and sizes.
Rinse Repeat is built around Photoshop, and the resulting changes remain editable because the system recreates them with native Photoshop tools. That detail matters for workflows where a creator may want to adjust the automated result afterward.
Experiments point toward a different creative workflow
Adobe calls these demonstrations “sneaks”: technology previews that might eventually become features in Creative Cloud apps or the Firefly platform. There is no guarantee that any particular project will launch. The article describes most of the experiments as built around Adobe’s generative AI work and says some could eventually reach users.
A common thread is making tasks in photo editing, filmmaking, and audio mixing feel more accessible and interactive. Some demonstrations may look playful, but each connects an approachable interface to a practical change in a creative project.
As senior principle Adobe research scientist Gautham Mysore told The Verge, “Making these interactions more intuitive is also about giving experienced creatives more precise control.” He pointed to Sound Bender’s sound exploration and Sight Span’s camera positioning as examples of how professionals could use their judgment to refine results.
If the experiments become useful products, they could change how creators approach work now done with editing tools or stock libraries. They also show why AI’s role in creative work is complicated: easier controls may help people shape results, while effective automated tools could also replace parts of existing workflows.