Adobe is working on generative AI video manipulation

Adobe is showing off a new generative fill feature, Project Fast Fill, that can easily add or remove objects in videos with the power of AI. It’s one of several new, wild, experimental AI features announced today at the company’s MAX conference. Project Fast Fill has the ability to swap in clothing accessories on people in motion or remove tourists from the background of a landscape pan. Fast Fill works similarly to Google’s Magic Editor, where you can move around people or objects — but with video instead of just still photos. Fast Fill promises to do things the company’s Project Stardust does for still photos, like change colors by just entering a text prompt. Adobe says these generative AI features are made possible by the company’s updated Firefly AI models. Like most new features presented at Adobe’s conference, Project Fast Fill is an experiment that the company’s engineers are working on for now. But like Photoshop generative fill features revealed in previous years, they often eventually do make it to Creative Cloud customers. Adobe’s eyeing the new feature for use in Premiere Pro and After Effects but has not provided a timeline. Adobe’s working on its AI editing tech for video, audio, and 3D design in other ways, too. The company’s Project Dub Dub Dub can translate voices into different languages. Scene Change can move subjects to a new location in another clip. Res Up uses diffusion to upsample lo-res video. And Project Poseable can pose 3D objects based on photos of real people and can generate 3D renderings using text input. As more AI generative tools for photos become more accessible to users, like Google Pixel 8’s Magic Editor, we were already at the stage of wondering, What is a photo? Now, Adobe’s AI video-manipulating Fast Fill could bring that same ability to more people soon, and we will have to ask: What is a video?