Originally Posted by AdultKing

With NVIDIA's face synthesis AI you can generate as many faces as you like, all different, once the model is trained, set up and running.

As technology develops and this becomes possible for video, speech and real time interaction the cost becomes no greater than the platform upon which it runs, because you will have infinite generational capacity from the AI.

Once this stuff becomes real and not simply in the realm of research then you won't need to get $3 - $4 per minute. The companies that successfully bring these things to market will only need to get cents per minute (I don't even think per minute charging would apply - it would be another billing model).