Looks good (actually terrible) and will be even better.
UPD:
Runway Gen-2
From text-to-image to text-to-video? // but not so easy, image-to-video is more realistic now..
The pipeline is text-to-image[s]-to-video[s]
It’s a patch, not a solution.. (but how to do better? No one knows)
A yet another example, just a sequence of “animated” images…
Boom, but how to do it differently?
Create characters, objects, backgrounds, effects, etc, like in 3D but not actually 3D, and compose all together in a scenes and sequences.
The combination of generative objects…
More example videos (ugly)
In a couple of years, text-to-video will be as easy as prompt to LLMs now! But now it’s in a fake it till you make it stage..
More examples..
And 3d engines…
