Adobe's Project Morpheus uses AI to automate frame-by-frame video edits

Something Adobe does at nearly every one of its annual design conferences is preview experimental features that may one day make their way to its apps. In 2020, it was a tool called Sharp Shots. This year, it was Project In-Between and a feature called Project Morpheus.

Building on the Neural Filters Adobe debuted last year, Project Morpheus is capable of automating frame-by-frame video edits in a way that produces consistent results. In the demo the company showed off, one of its employees modified a video of themselves where they had a neutral expression throughout to put a smile on their face. They did it in mere moments, with Adobe’s Sensei AI most of the heavy lifting. You only need to see the demo once to understand the potential for misuse, something Adobe told Engadget it's taking into account. 

“As with all Adobe Max Sneaks, Project Morpheus is a way in which we preview exploratory and forward-looking technologies from our research labs and engineering teams. These proof-of-concept ideas aren’t always destined for inclusion in products,” a spokesperson for the company said. “We recognize that we have to balance innovation with our responsibility to ensure our technology is being used for the good of our customers and society. Our development of AI is guided by the principles of Accountability, Responsibility and Transparency.”

Additionally, the company pointed to its work on the Content Authenticity Initiative. First announced in 2019, the project has seen Adobe work with The New York Times and Twitter to develop a metadata tagging system designed to limit the number of altered images and videos that circulate online. At Max 2021, it also started beta testing Content Credentials. It allows professionals, including photojournalists and artists, to attach attribution data to their images that detail the edit history of a file.