First frame
Last frame
JPEG, PNG or WebP, up to 9MB each.
optional
Your transitions will appear here
Fill both frame slots on the left — the generated motion between them lands on this canvas.
Give the model where the shot starts and where it lands. It invents everything in between.
Every other method asks the model to invent an ending. Here you supply it, so the clip cannot wander off somewhere you didn't want.
Each slot is labelled and held separately, so uploading the ending shot first never turns it into the opening one — a real hazard of plain multi-image uploads.
Models that accept just one frame are filtered out, including two that would quietly ignore your ending shot rather than fail.
Three steps, no setup
Where the clip opens.
Where it lands. Both are required before generating.
The model builds the motion between them; add a prompt to steer the travel.
You give the model two pictures — where the clip starts and where it ends — and it invents the motion in between. It's the most controllable kind of AI video, because you've fixed both ends of the shot.