Extension is harder than generation
A new clip starts from nothing. An extension has to match what came before. This model reads up to ten seconds of your footage first.
What stays the same
Character identity, lighting and framing carry over. That is the whole point of extending rather than generating a second clip.
Four ways to continue
Carry on, loop back to the opening frame, slow the motion out, or pull the camera back to reveal more.
FAQ
Can it extend video with AI from any source?
Any clip you own works. Clean footage with steady motion extends most convincingly.
How long can the extension be?
Four to ten seconds per run. Run it again on the result to go further.
Is this like Grok extend video?
Same idea, different engine. Here the model reads ten seconds of context, not just the final frame.