I think AI is moving from trying things for fun toward building actual workflows.
Seedance 2.5 on September 24 was a good example.
It cost more per second than 2.0, had no fast tier and no 4K output. But it could fix a specific moment in a clip. If there was a hand glitch at second 11, you did not necessarily have to regenerate the entire sequence.
You could draft at 480p and only render the version worth keeping at 1080p.
For creators, cost per second does not tell the whole story. The number of generations needed to get a usable clip is part of the equation too.
That is what connecting tools into a workflow looks like to me, rather than simply adding another generation button.
#SHELL #MyShell
Seedance 2.5 on September 24 was a good example.
It cost more per second than 2.0, had no fast tier and no 4K output. But it could fix a specific moment in a clip. If there was a hand glitch at second 11, you did not necessarily have to regenerate the entire sequence.
You could draft at 480p and only render the version worth keeping at 1080p.
For creators, cost per second does not tell the whole story. The number of generations needed to get a usable clip is part of the equation too.
That is what connecting tools into a workflow looks like to me, rather than simply adding another generation button.
#SHELL #MyShell