just post everything in one thread, its crazy
it made ONE. ONE photo of mona lisa start talking
There was a good deal of multifaceted AI banter in this NVIDIA topic (how machines learn, AI generated music sources, etc), so I figure it's relevant to link it here as a callback.
it made ONE. ONE photo of mona lisa start talking
It's not the best replication, but getting this many angles out of one shot is insane. At this point it's just a matter of refining it's accuracy, which is easier than what's already been accomplished.
I've looked into (and experimented with) Deepfakes for a little bit, and that stuff takes a ton of compiling time and two seperate lengthy video sources to gather composites of their faces for mapping one onto the other. If this goes further, then mapping people onto pornstars and stuff should become startlingly easier.