It converts an open-ended prompt into a bounded problem. Ordinary image-to-video knows where it begins and improvises the rest; give it a destination as well and the model has to land there, so the arc of the shot is decided by you rather than discovered by the model.
That makes two things easy that are otherwise fiddly. Loops: make the last frame identical to the first and the clip cycles seamlessly. Transitions: make the last frame of shot one the first frame of shot two and the cut disappears into a generated morph.
The two frames have to be plausibly connectable in the time available. A model asked to travel from a close-up interior to a wide exterior in a couple of seconds will find a route, but the route will be a warp rather than a camera move.
In practice
- Same image in both slots gives you a loop with motion in the middle.
- Keep the two frames in the same location, lighting and lens family or the interpolation reads as a morph.
- The prompt still matters — it decides how the model travels between the two, not just that it does.
The mistake to avoid
Choosing an end frame that is compositionally unrelated to the start. The model will connect them, but the connection is the artefact everyone notices.
Go deeper
First/Last Frame: Perfect Transitions and Seamless Loops
First/last frame AI video control explained: seamless loops, invisible transitions between scenes, morph reveals, and frame-pair design rules.
Where you will run into it
- The Sweet Switch: Fruit Heist — A high-stakes 3D Pixar-style heist movie featuring a squad of fruits trying to sabotage a salad, only to realize they are the target for a smoothie.
Related terms
Image-to-video
Image-to-video animates a still you supply: the picture becomes the opening frame, and the prompt describes only what happens next.
Video extend
Video extend continues an existing clip past its final frame, generating new footage that starts from where the old footage stopped.
Reference image
A reference image is a picture supplied alongside the prompt so the model can copy an identity, product or style from it, without that picture becoming a frame of the output.
Frame interpolation
Frame interpolation invents new frames between existing ones, raising a clip's frame rate or slowing it down without it becoming choppy.
Seed
A seed is the number that decides the random starting noise for a generation, so the same seed with the same settings reproduces the same output.
The all-in-one AI studio for creators. 60+ models for video, image, voice, music and lipsync in a single app.