[–] 18853274? 0 points 2 points 2 points (+2|-0) ago
Generate a dataset of deepfake generated faces with many different techniques, generate a dataset of known non-deepfaked faces, train another neural network to detect if face is generated or not.
Back in reality, corroborating evidence breaks down the 'Wuznt me I dindu ' bullshit.
[–] 18853550? 0 points 4 points 4 points (+4|-0) ago
DeepFake's take a decent amount of time right now to render. Any decent quality one especially takes more time.
That said most images leave digital fingerprints of animation engines and the like. Forensics will be able to to detect what is real and what isnt for some time until the AI gets so cheap and perfected we genuinely won't know what the fuck is real and what isn't,
For now - this is a lame attempt to keep the fooled, fooled. "Don't worry that's not real." says the trustworthy news source. "Don't go looking any further. It's just fake. Go back to watching Ellen. Here's American Gladiators if that's not enough. Wow the Kardashians did WHAT?"
It won't permeate well and if what Q+ provides is a steady stream - there is no defense. They won't be able to label hours and hours of footage as false. And when Courts uphold their validity the DeepFake narrative goes immediately out tit's short-sided and desperate window.
[–] 18853234? ago
https://files.catbox.moe/zrue2o.png :
This has been an automated message.