Spring 2018: GenmoGenmo replaces boring, impersonal photo filters with generative AI effects
Jason Salavon, Associate Professor, The University of Chicago Department of Visual Arts, Logan Center for the Arts; Faculty and Fellow, The University of Chicago Computation Institute.
Genmo (Generative Mosaic) is a neural network driven visual effects app that recreates any photo or video using an entirely separate set of images. With Genmo, users can upload a video, select a dataset to render it with, and get a recreated clip back in seconds. It’s AI-powered creativity for your phone.
Substantial portions of the world’s population are participating in social photo and video sharing every day, a trend that continues to see growth every month. In an age where we pursue endless fascination and variety, prevailing social media provide unintelligent photo and video filters that are both are boring and impersonal. Filter fatigue requires that popular services, like Snapchat, constantly provide new filters because of their inherently short shelf-life.
Genmo, perhaps the first artificial intelligence sandbox for your photos and videos, aims to engage the creative power of millions of social media users to provide an innovative suite of visual effects that continuously generates new and surprising outcomes easily shared on the most vital social networks. The proliferation of user generated content and the creative limitations of existing technologies have paved the way for artificial intelligence to rethink the social photo/video creation and sharing experience, allowing for content creators to leverage their idiosyncratic behaviors and augment their visual production.
Developed by LatentCulture, a research group focused on generative systems and AI for creative applications, Genmo’s core artificial intelligence technology is patent pending and allows for the reconstruction of any arbitrary source image or video with another image dataset. This produces an output that is truly generative and not limited by the simplistic constraints that define current social filters (facial mapping, stock overlay images, etc). The key feature of the service we are creating is that it will continually incorporate new visual effects and modification abilities as they are produced and perfected by our team, improving upon the rotating carousel of filters that Snapchat offers by handing over more creative agency to our users. As our technology begins to learn about our users’ content interests and behavior, the effects and dials may evolve uniquely for each user.
Jason has over two decades of experience as a practicing artist using generative and data-driven technology to create his work; He takes complicated technical processes and translates them into compelling, expressive visual experiences. He leads the group LatentCulture at the University of Chicago, which explores the creative potential of generative systems and AI. With an emphasis on real world production, LatentCulture builds deep neural networks to engage with the problems and opportunities of creating original art through computational means.
How does the App Challenge help with this great idea?
The LatentCulture team is looking to further develop and implement an effective UI/UX for Genmo that allows for access to their technology and the ability to adjust parameters of the effects. Although a number of effects are ready for immediate implementation, the team is looking to build an app that can accommodate an ongoing series of updates and new features.
How does this support UChicago's research mission?
Artificial intelligence technologies, while they sound “cutting edge”, are largely opaque and difficult for those not already working with them to understand. Genmo would allow masses of social media content creators to engage with AI technologies in a progressive and deeply personal way, promoting an affinity for how the technology works (around datasets, training, iteration, and generative results).