Have you ever surfed social media, seen something cool, and then thought to yourself – WAIT – let me try that, I think I can do that.
That’s me all the time. Rewind back to late 2020, early 2021. We were all settling into our work from home, COVID-19 lives. Prime time for tinkering technologies.
Our Chief Innovation Officer and Chairman of Futures, my boss, Chris Perry had an upcoming anniversary, 21 years at Weber Shandwick. No time like the present I thought to make a deep fake of him.
I honestly didn’t know where to start. I knew of software for CGI and other things that could render from scratch, but I wanted to know if I could create this using low-or-no-code approaches.
Some googling and VOILA!
I found Descript, a platform that does transcriptions of video content and, also had a voice learning model. And so one could manipulate the transcript and re-order the words and sounds – and apply it on top of Descript’s learned voice of Chris Perry. BOOM. I made a new soundtrack – no code, just configuring and re-ordering of the words and sounds. I thought I was a DJ. In full transparency, it did take me about a day to configure and learn Descript, it’s fairly straight-forward, but manipulating the content – as well as input content into the platform required a few re-starts.
Further searching and I found some online open source early-stage models that would allow me to mash the new soundtrack and a course video – to manipulate the video and generate a new video – these were academic testing – available via web from University of Michigan, as well as European universities with models to test.
– Patrick Chaupham