PROMPT GUIDE

How to Make AI Video with a Start and End Image (tips for both frames)

Some video generators have a feature where you give a first image and a last image and the AI fills in the part between. Used well, it lets you create changes in face direction or expression the way you want. This guide covers how to choose the two pictures, how to write the motion and how to fix it when the middle breaks.

Last updated: 2026-10-22 approx. 3 min read

日本語版

The short answer

When specifying a first and last image, match the background, outfit and composition, and change only one thing, such as the face direction or expression. The smaller the difference between the two, the less likely the middle is to break, and the prompt just needs a short line on how it changes.

What is AI video with a first and last image?

You give two pictures, a start and an end, and let the AI fill in the motion between them.

Method	What you give	Good for
First image only	One picture plus a line of motion	Easy animation, moving hair or the background
First and last images	A start picture, an end picture and a line of motion	Creating a change of face direction, expression or scene on purpose
Text only	Text (a prompt)	Making a scene from nothing, with no picture

Whether this feature exists, what it is called (start, end, keyframes and so on) and how many times you can use it change by service and plan. Check the official guidance before using it.

Anime girl looking away toward a window, a start frame for a start and end image video — 最初の絵（横を向く）

The same girl at the same window smiling toward the viewer, an end frame for a start and end image video — 最初の絵（横を向く）

Illustration usage rules

How do you choose the start and end pictures?

Match the background, outfit and composition, and change only one thing, such as the face direction or expression.

Match these	These can change	Reason
Background and place	The direction of the person	If the background changes, the scene smears and breaks partway through
Outfit and hairstyle	Expression	If the outfit changes, it melts into a different outfit partway through
Composition and size	Hand position (slightly)	If the sizes differ, the camera moves on its own
Direction of the light	Gaze	If the light differs, the brightness flickers partway through

Tips for matching the face across two pictures How to Keep a Character's Face and Outfit the Same in AI Video (consistency tips) →

Steps for making a video with a start and end specified

Prepare the start picture (for example, looking out of a window, turned sideways)
Prepare the end picture in the same place, outfit and composition as the start, changing only the face direction or expression
In the video generator, choose a mode that lets you specify both a start and an end, and load the two pictures
Write how it changes in one line in the prompt (for example, she slowly turns toward the camera and smiles)
Make a few and compare them, and choose the one where the face and outfit do not break in the middle

How do you write the motion? (prompt examples)

The two pictures almost decide the answer, so write only the change and the camera, briefly, in the prompt.

The girl slowly turns from the window toward the camera and smiles softly. Static camera, soft daylight.

Her expression changes gently from calm to a warm smile. The camera stays still.

The camera slowly pushes in as she looks up and meets the viewer. Calm mood.

How do you fix it when the middle breaks? (common failures and fixes)

Failure	Cause	Fix
The face distorts in the middle	The face direction differs a lot between the two pictures	Reduce the difference in direction and split it into two steps
The outfit or background melts	The outfit and background do not match between the two	Remake the two pictures with the same outfit and background
The motion is awkward	The video is too short, or the difference is large	Reduce the difference, or choose a longer setting (check the official guidance)
It does not reach the end picture	The motion in the prompt is too strong	Shorten the prompt and write only the change

What are example start and end combinations?

Start with a combination with a small change, and once you are used to it, make the change larger to reduce failures.

Start	End	Difficulty
A neutral face	A smile	Easy (the face direction stays the same)
Looking sideways	Looking at the camera	Moderate (keep the difference in direction small)
Standing	Waving	Somewhat hard (the hand moves)
Standing small in the distance	Coming closer and appearing large	Hard (the camera move is large)

For a hard combination, adding one picture in the middle and splitting it into two shots makes it more likely to work.

Rinka's tips

Line the two pictures up side by side, and if the position and size of the face are nearly the same, the motion in between tends to connect naturally
In the end picture, not changing the shape of the hands or hair much from the start makes it less likely to break in the middle
Note down the pictures and prompt for a combination that worked, so you can reuse it for the next shot
First try a simple combination where only the expression changes, and once it works, move on to changes of face direction
To join three or more pictures, make them as separate shots two at a time and line them up at the end
Making the end picture by changing the start picture slightly as a still helps the outfit and background match

The basic steps for animating with only a first picture How to Use Image-to-Video AI (Steps to Animate a Still Image) →

How to use the end picture as a joint to make a longer video How to Make AI Videos Longer (Extend, Chain, Last Frame) →

How to loop a video by using the same picture for start and end How to Make an AI Video Loop (Tips to Hide the Seam) →