Went to a deep fake lecture in Austin and left with more questions than answers
I drove down to Austin last weekend for a talk at a tech meetup about spotting AI-generated video. The presenter showed a bunch of classic signs like weird ear shapes, no blinking, and unnatural skin blur. But here's what bugged me: everything was focused on old models from like 2022. They still push the 'look for bad teeth' method, and honestly, in the last 6 months I've seen fakes that pass all those checks. I tested one on my own laptop with a free tool and it fooled the metadata scan completely. The community seems stuck on these outdated tells while the people making this stuff are already way ahead. I work in IT so I see the newer stuff come across my desk, and it's scarier than what that guy was showing. Has anyone else noticed the detection guides on here are falling behind what's actually out there right now?