I recently had to give a short organization wide presentation about myself. My initial reaction was to opt out.
Upon second thought, it might be a good opportunity to code and test out Deep Live Cam on a desktop with a NVIDIA GeForce RTX 4080.
The Frames-Per-Second (FPS) ranged around 10fps to 15fps when a face was detected. The M2 Macbook Air was getting around 3-6fps.
With the consent of a co-worker, I was able to use his likeness for the ~10min presentation.
The presentation was conducted though a custom terminal written in HTML/JS/CSS that is now part of my Github profile. You can access it here. I left some CTFs and Easter Eggs and probably add to this over time.
The content of the interactive terminal is parsed from the readme.md. GitHub Pages will not include the markdown file, so I added a git pre-commit hook to base64 encode the markdown and store it as a JavaScript constant variable. The details of the script are found here. The real-time deep fake was revealed at the end of the presentation.
It is a fun game to load up a deep fake of a public figure or celebrity and have your friends guess who you’re portraying.
Previous post on Deep Live Cam.