It is often said that a picture is worth a thousand words; we believe that it's worth a thousand videos. When Shazam came out, it redefined the difficult task of finding a song based on a tune - we thought it'd be awesome if we could similarly, change a picture into something fun! Each photo has a story to tell, and we wanted to bring it to life.
What it does
PhotoVine, cleverly (or not? :P) a portmanteau of Photo and ToVine converts a photo to a playlist of Vines based on relevant themes, giving each photo a unique personality told within a a series of around 30 seconds. Using Vine and Clarifai's powerful APIs, PhotoVine not only finds clips with similar themes as a picture, but even analyzes each Vine and filters results based on actual video data!
Bored? PhotoVine a pic - you might unlock something neat.
How we built it
Along with a whole bunch of Kickstarts and (maybe) a few rage-quits, PhotoVine began as a simple web application that polled information using Clarifai's image tagging services. Each of these tags are then fed into Vine to get clips that are somewhat related, and a small sample of victims -er, we mean potential clips are taken to be analyzed one last time for actual content relevancy by Clarifai. What remains is a small playlist of most relevant clips for the user to enjoy~
Challenges we ran into
Originally, we had thought of using Clarifai's API in conjunction with Youtube, but Youtube has terms that prevent its videos from being downloaded, making no way to get them to Clarifai. We even tried to rip videos off using some cough rather hack-ish cough methods (but hey, this /is/ a hackathon, right?).
Accomplishments that we're proud of
We're more than happy that everything came together nicely in the end - we were worried
What we learned
With a bit of creativity and innovation, you can really find fun results from even something as analytical as image recognition; it's all about the direction you view the data from and we believe PhotoVine's not only something amusing, but a pioneer for accurate theme-based photo-to-video searching.