Help me analyze/understand the background of this AI video?
Hi, so I've been thinking about this for several days now, and thought it might be an interesting topic for Tildes.
Earlier this week, YouTube suggested this AI Sitcom video to me. Some of the jokes are actually very cohesive "Dad jokes", and it got me wondering how much of the video was AI generated. Are the one-liners themselves AI generated? Was this script generated with AI, and then edited before passing it on to something else to generate the video and voice? Or are we at the phase where AI could generate the whole thing with a single prompt? If it's the latter I find this sort of terrifying, because the finished product is very cohesive for something with almost no editing.
I'd also be interested in discussing where this video might have come from. The channel and descriptions have almost no information, so it seems like this may be a channel that finds these elsewhere and reposts? Or maybe the channel is the original and just trying to be vague about technology used?
Also side note, I have no idea if this belongs in ~Tech, so feel free to move it around as needed.
Google veo 3 makes these. All the ones of a person doing a vlog style are from the as well, you can start to see the look and feel of it after a few. Veo I think makes good use of massive server farms for the amount of quality it can put out. Not sure if the script was generated or part of the prompt though. I believe veo does the lip sync and voice gen though.
In some instances the voices are crazy good. For example, at this point in the video, if I close my eyes I could swear it's Paul Rudd speaking.
What bothers me is that I can see how well it copies the format, and if the barrier to entry is low enough we're headed for a future with a lot of rampant misinformation. Actually, it's not even really a question of "if" or "when" in my mind. It's about how big the problem will be. With enough horsepower, these videos could be cranked out a rate that could overwhelm our ability to filter or detect. In some ways I think we're already at that point.
I saw some other videos (that I can't find again at the moment) where a news reporter lady is seen covering natural disasters. The videos were very convincing, but thankfully included ridiculous premises in some cases and in a few they ended with the reporter saying "Just kidding I'm not real". I think the artist's intent was to draw attention to how dangerous this tech could be, but it's easy to imagine it actually being used to deceive on a massive scale. Couple that with our short news attention spans, and I could easily see a video making the news and being accepted as fact before a follow-up could be made to debunk it.
I'm really trying to temper my fear here.
Edit: Found an example of the fake reporter video
Edit2: And that video led me to this Veo 3 compilation
I promise this is my last edit: AI Reacts to Being AI: Google Veo 3 Test Footage