[kitx]

One can imagine that resolving this will require a separate model to construct a timeline/keyframe series from a text description then interpolate between them.

Fascinating times, either way this is allegedly being open–sourced, so I'm hopeful others will be able to build on top.

[lairv]

I wouldn't have thought we'd have these text-to-video models before the end of this year, so we might actually not be that far from minutes-long video generation