Mobile App Uses AI Technology to Edit Short Social Videos

Social video startup TRASH aims to make it simple for users to create short videos to share with friends. CEO Hannah Donovan, previously general manager at Twitter’s now defunct Vine video app, co-founder of music-discovery site This Is My Jam, and former head of creative at Last.fm, points out that research shows most consumers carry powerful cameras with them, but don’t know much about editing video content. So she set out to leverage computer vision technology to analyze video and synthesize the content into an appealing sequence. The free TRASH app does just that, as explained by its simple tagline, “You shoot, we edit.”

Currently in creator beta, the iOS app generates an edit of uploaded video footage using technology developed with a National Science Foundation grant. The app’s AI creates a rough cut and the user has the option to reorder clips and make changes to the pacing, music and visual effects. Donovan is inviting creators to test the app and provide feedback.

“TRASH isn’t just about putting filters on top of your video, it’s also about helping you generate a sequence. Today’s short videos with soundtracks are just the very beginning,” explains an introductory blog post. “We want everyone to be able to participate in the medium of video, democratizing a medium that was previously only for pros.”

Donovan recently demonstrated the app for TechCrunch, shooting some short clips and assembling them into a sequence. According to the reporter, the result was, “not exactly an amazing video but much, much better than anything I could have done with the footage. We also got to tweak the video by adjusting the music, the speed or the ‘vibe,’ then post it on TRASH and other social networks.”

The technology behind the app uses analysis and synthesis to create videos. With analysis, “a neural network analyzes the footage to identify elements like people, faces, interesting actions and different types of shots,” notes TechCrunch.

With synthesis, “we try to figure out what are the most cool and interesting parts of the video, to create a mini-music video for you with a high diversity of content,” explained Donovan, adding that the app should get smarter with more training data.

Donovan founded TRASH with chief scientist Genevieve Patterson. The startup has raised $2.5 million so far from the National Science Foundation, Digital Garage and Dream Machine. There are plans to eventually build an Android version of the app.