OpenAI's Sora video generator is impressive, judging from the examples Sam Altman is posting to Xitter. There's no release date yet for Sora, but even now, anyone can easily create pretty realistic short snippets of video using Stable Diffusion.
Take a look at this video, which shows dozens of examples of fake webcam videos. The creator wrote:
Inspired by the recent financial crime incident, where hackers held a video meeting with AI generated video of employees, I decided to make a video call-esque type generation to see how convincing it could be. There are better ways to do this but for speed and resources i just used stable diffusion, with SVD and flow frames to produce short video clips ranging from 2 to 14 seconds in length which i then pieced together. The output is not that impressive if your up to date on the latest technology, but still shows what can be achieved, if you had shown me these videos a few years ago, i wouldn't of though it was possible. If you can tell that most of these videos are fake, then wait a couple months, if you cant tell, then welcome to the future.
As a bonus, the narrator of the video is an AI voice.
See also: "The Office" but Pam is everyone
