Inspiration
This make-a-thon is all about using technology to create art. Together, we decided we could combine both our separate passions in art (music and videography) to create something that represented how we have to take risks to continue our passions. This project was all about taking a risk and jumping into the deep end with what we wanted to do, and we wanted our video to reflect that.
What it does
This video combines a lot of technological elements to help convey the message of choosing your passions over a job. We start off the video with cutscenes from pop culture that represent characters about to embark in a life-changing journey. Throughout the video, an AI visualization of our song can be seen, which uses a GAN to create images based on our song input. This helps create the overall theme of a drug trip, representing how different and more full life is when you actively choose to pursue your passion without fear. Then, we cut to shots taken around Georgia Tech and deep fakes, which also used a GAN to generate lip poses to synchronize with the audio. The live shots were edited to also continue on with the theme of life being different with passion.
The song itself revolves around getting past the fear of pursuing your passion, which has been the hardest step for both of us. Overall throughout this project, there was a lot of letting go of our inhibitions and what we were afraid to do for the sake of what we enjoyed.
How we built it
The song was built using GarageBand, FL Studio, and Audacity. The video was created using Adobe Studio. The visualizations in the background of the video were created using a GAN that took the music as input and created frames. The deep fakes were created using Wav2Lip
Challenges we ran into
Deep fake models sometimes have a hard time recognizing where the lips are in videos, especially with cartoons or when multiple people were present. There was a lot of tweaking the model so that we could ensure that it would detect the right features in the videos.
Accomplishments that we're proud of
We are really proud of the elements we were able to stitch together in the video, especially with the results of the visualizer and how well it fits into our theme of tripping into Wonderland.
What we learned
We learned a lot about how deep fake technology works and how various groups are trying to create models that can accurately detect them. The technology is powerful and, if used unethically, can definitely be dangerous without checks in place to detect them.
What's next for Wonderland - R&R
We would like to make the song a bit longer and work on creating some more visualizations to truly exuberate the "trippiness" of the video a little better.
Log in or sign up for Devpost to join the conversation.