logoalt Hacker News

swiftcoder • today at 10:47 AM • 1 reply • view on HN

> Do you just accept that we are now able to generate any image of everything we want? Video? 3D Models? This was unimagbinable a few years back, now you can generate live video 24/7.

And its still kind of shit? AI video in particular is stuck in an uncanny valley, because the video models have no internal framework for how light transmission works...


Replies

origini • today at 12:04 PM

I have seen plenty of AI video which is not stuck at all in any uncanny valley.

I'm also quite a stickler for light as Im doing photography for 2 decades.

I have seen shitty AI videos for sure and I still can spot AI images on IG but it got a lot harder.

➕ show 1 reply