First & Last Frame: keyframe your way to a perfect shot.

Text-to-video gives you a shot. First & Last Frame gives you the exact shot. You provide the opening still, the closing still, and a one-line description of the motion in between — the model interpolates the rest. It is the closest thing to keyframing in traditional animation, ported into the generative era.

For working creators this mode is the cheat code. You stop fighting the model for composition and start handing it the answer.

The mental model

Think of First & Last Frame the way an animator thinks of keyframes. You set two anchor poses, and the engine handles the inbetweens. The motion is constrained by your choices — which means it lands where you want it to land, every time, instead of drifting somewhere weird in the last second.

Where it wins

• Product reveals. First frame: closed box. Last frame: product hero pose. Motion: smooth reveal. Done.
• Match cuts. First frame: end of previous shot. Last frame: start of next shot. The model bridges the cut seamlessly.
• Logo animations. First frame: scattered shapes. Last frame: locked-up logo. Motion: assembly.
• Title sequences. First frame: blank space. Last frame: hero typography. Motion: reveal of your choice.
• Transformation shots. Day to night, calm to storm, raw to plated dish, before to after.

How to choose your two frames

The quality of your output depends almost entirely on how well your two frames pair. Three rules:

1. Same composition. Don't change the framing radically between frames. Keep the camera in the same place — let the subject move, not the camera, unless you specifically want a camera move.
2. Same lighting. If frame one is golden hour and frame two is high noon, the model has to work too hard to bridge the gap and you'll see banding or color drift. Match light.
3. Same identity. The subject in frame one should be the same subject in frame two. Use the same character anchor or the same product reference for both.

The easiest way to satisfy all three: generate frame one in Nano Banana Pro, then use the in-app edit to create frame two as a variation. Same model, same seed, controlled change.

Writing the motion line

The motion line tells the engine how to interpolate. Be specific about camera and subject movement:

"Camera holds. Subject [action]. [Speed/timing]."

Worked examples:

• "Camera holds. The runner crosses from frame left to frame right. Sustained mid-pace."
• "Slow push-in. The product rotates ninety degrees clockwise. Smooth, linear."
• "Camera holds. The candle flame grows from a spark to a full blaze. Three-second build."

"Once you start building shots from two frames out, you stop generating videos and start directing them." — AI Studio Editorial

Pro pattern: chained First & Last Frame

For longer pieces, chain multiple First & Last Frame jobs. The last frame of generation one becomes the first frame of generation two. You get a continuous sequence with rock-solid continuity, because every transition is anchored on a known frame.

This is how we build product films and short-form narrative pieces inside AI Studio. Six anchor frames. Five interpolations. One coherent ninety-second piece.

The bottom line

First & Last Frame is the mode you reach for when you already know what the shot needs to look like. It's faster than text-to-video for any deliverable shot and it puts the composition decisions back in your hands. If you've been bouncing off random outputs in pure text-to-video — switch to First & Last Frame and watch your iteration count drop.