PictureToVideo AI
Nobody filmed the two seconds between the photo of them apart under the veil and the photo of the kiss, and nobody ever will. Here is how image to video with start and end frame rebuilds them: the first frame slot takes the shot it opens on, the last frame slot takes the shot it must arrive at, and one required prompt line tells the model how to travel between them. The model then generates only that middle stretch — 480p to 4K, five to fifteen seconds — so the start and end frame clip moves from the first photo to the second instead of cutting between them like a slideshow. Animating a single photo is a guess about where the motion goes; give it both start and end frames and the destination stops being a guess, because you already decided it.
Both ends of the start and end frame clip are yours, and the middle is one required line. Nothing about where it lands is left to the model's taste.
The first frame slot takes the photo the clip opens on; the last frame slot takes the photo it has to reach. Each slot holds a single image — no third slot, no way to stack a longer sequence — so the start and end frame pair is the anchor for the whole run.
Both slots cap at a single image. The start and end frame pair is the anchor, not a timeline.
It cannot drift past the shot you set as the end frame. That is the point.
Swap which one sits in which slot and the same pair runs in reverse.
The frames fix where the start and end frame clip starts and stops; the required prompt decides how it travels. "She leans in and the veil lifts" lands the same place as "the camera pushes past her toward him", but the middle plays out differently. The field takes up to 20,000 characters, though one sentence usually beats a paragraph.
The two start and end frames already say what. The line only has to say how.
Say which of the two start and end frames moves, or the model will choose for you.
Keep both frames, rewrite the middle, and run the pair again.
Resolution, length, aspect ratio and sound are all settled before the start and end frame run. 480p through 4K, five to fifteen seconds, seven aspect ratios including Auto, and an audio switch — Seedance 2.0 renders all of it in one pass.
Four resolution steps, from a quick test to a sharp 4K master of the same start and end frame pair.
Three clip lengths, enough to carry a moment from the first frame to the last.
Seedance 2.0 can generate sound under the clip on the same run — no separate audio step.
The controls behind every start and end frame clip — the frame lock, the one-line steer, and the Seedance 2.0 engine that renders the two ends into one.
The start and end frame pairs that get uploaded once you stop asking a single photo to guess where it was going.
The photographer caught them apart beneath the veil, then again mid-kiss, and nothing between. A bride sets that start and end frame pair — the first shot as the opening frame, the kiss as the closing one — writes "she leans in slowly as the veil lifts", and gets back the two seconds the album never held.
A creased print of the family house in 1974, and a phone shot of the same porch this spring — a single start and end frame pair. The old photo goes in as the start frame, today's as the end, and a line like "the walls weather and the garden thickens over the years" sets the drift. A first and last frame ai video then travels the fifty years instead of jumping them.
One photo of the empty flat on handover day, floorboards and nothing else. One of the same corner after the sofa, the lamp and the rug arrived — the two ends of a start and end frame clip. The empty shot leads, the furnished one closes, and a line like "furniture and warm light settle into the room" guides it. Start frame end frame ai walks the room from bare to lived-in without a single cut in the middle.
Two photos and a sentence. The order matters less than the start and end frame pair you choose.
What people ask once both ends of the start and end frame clip are their problem rather than the model's.
Discover our full suite of AI-powered creative tools
The one it starts on, the one it lands on — your start and end frame pair — plus one required line for the seconds between. A Basic plan opens the run; you already own the hard part.
0 / 20000
Video generation is available on Basic plan and above.
Pick a version, describe your video and press Generate.
Sample output