>>29149850You fundamentally misunderstand how the technology works.
There is no "improvement." They set the quality by how they make the model. To scale up takes exponentially more resources, and they are already burning tens of millions of dollars per training exercise. It would require tens of BILLIONS to improve the extremely limited model to higher fidelity. You simply have no fucking clue what you're talking about. For that amount of money, you could hire ten thousand people for their entire lifetime, from birth to death. Which one will improve faster, retard?
Focus on the fact that you've never seen a perfect deepfake: They only get differently retarded.