Yeah. Just a few years ago, people here would have said stuff like that was decades away at best and pure science fiction at worst.
Yeah. Just a few years ago, people here would have said stuff like that was decades away at best and pure science fiction at worst.
I remember having to tell people "Don't describe too many individual objects in your DALL-E 2 prompt, because the model has trouble separating discrete concepts and the features tend to blend together."
Now we have photorealistic video with sound and, oh yeah, the model can generate an entire script and mini-plot on its own based on the most basic prompt.