AI
Adobe Max Showcases AI Experiments: Photos Into 3D Scenes, Characters Shaped Through Conversation
Adobe announced several experimental features at its Max conference, including Project Screen Test for designing AI characters through interview-style conversations, Project Sight Span for expanding flat photos into walkable 3D spaces, and Project Sound Bender for generating and adjusting sound effects with text and a sound canvas.
According to The Verge, Adobe showcased a batch of experimental “sneaks” at its annual Max conference. Project Screen Test is aimed at AI video character development: users first set base personality traits such as anxious, shy, or clumsy and choose a voice through presets, text descriptions, or uploaded samples; they can then talk with the character like a FaceTime call to fill in backstory, temperament, goals, flaws, fears, and mannerisms.
After the character setup is complete, users can describe an environment or scenario and have the tool generate an animated audition. The Verge认为 the results are more distinctive and consistent than many personality-deficient AI animated characters, but also noted that the project is still a standalone creative exercise and it is unclear whether or how it will enter Adobe’s existing AI video tools.
Project Sight Span can reconstruct flat images such as photos, paintings, or illustrations into navigable 3D environments with changeable camera angles, and generate objects and backgrounds beyond the original frame in a coherent style. The report noted that images may become pixelated and outlines unclear during navigation, but resolution recovers once an angle is selected, and details such as tattoos and facial features remain highly recognizable.
Project Sound Bender is aimed at AI audio mixing. Users upload a video and describe the desired sound effect in words, such as “bubbling sci-fi cauldron sound,” and the tool generates and automatically syncs the effect; different corners of a square sound canvas provide related AI sounds, allowing users to adjust sound characteristics in real time.