A keyframe is a frame where you fix exactly how something should look or where it should be at a set moment, and the software works out the frames in between. In AI video, keyframes usually mean the start and end images (and sometimes middle ones) that you give the model to move between.
Think of planning a road trip by choosing only the stops: breakfast in one town, lunch in another, dinner in a third. You do not plan every bend of the road; the map app fills in the route. Keyframes are the stops. The software, or the AI, drives the road between them.
In a sentence: “I set a keyframe with the door shut and another with it wide open, and the tool animated the swing in between.”
What does keyframe mean in animation?
The plain point: you set the important moments; the computer fills the gaps.
In animation software, a keyframe records a value, such as a position, a size or a colour, at one point in time. The manual for Blender, a 3D animation program, defines it in one line: “A Keyframe is simply a marker of time which stores the value of a property.”1
Set two keyframes and the program calculates everything between them. Blender’s example moves a cube between two keyframes and says the program “will automatically determine the correct position of the cube for all the frames between” them. How that motion feels (steady, easing in, bouncing) depends on the interpolation method you choose, which is the rule for filling the gap.1
What does keyframe mean in AI video?
The plain point: the keyframes are pictures, and the AI invents the motion between them.
AI video tools borrowed the idea, but the “value” is now a whole image. You give a start frame, an end frame, or both, and the model makes a clip that travels between them. Google describes this for its Veo video model: “By providing a starting and an ending image, you can direct Veo 3.1 to generate the transition between them.”2 Google’s cloud documentation names four models that take a first and a last frame: Veo 3.1, Veo 3.1 Fast, and previews of Veo 3.1 Lite and Gemini Omni Flash.3
There is one big difference from classic animation. Blender calculates the in-between frames by a fixed rule, so the result is predictable. An AI model imagines them, so the path between your keyframes can surprise you, and a second run may differ. The prompt matters: describe the movement you want between the two pictures. First frame and last frame in AI video shows how.
Keyframes let you pin down where a shot starts and ends. In AI video, make your start and end images match in style, light and characters, and use the prompt to describe the motion between them.
Why it matters to you
- Control: a plain prompt leaves the ending to chance. An end keyframe fixes it, which helps when one clip must cut neatly into the next.
- Planning: designing the key images first, then animating between them, is one way to plan a film. Plan it before you generate it covers working this way.
- Loops: giving the same picture as first and last frame can help a clip end where it began; see seamless loops.
Often confused with
| Word | How it differs |
|---|---|
| Frame | Differs: any single picture in a video. A keyframe is one you set on purpose; the rest are filled in. |
| Image-to-video | Differs: animating from one starting picture. Using both a start and an end keyframe adds control over where it finishes. |
| Storyboard | Differs: a planning sheet of sketches for a whole film. Keyframes are the exact images fed to the software for one shot. |
Questions people ask
What is a keyframe in AI video?
It is an image you supply for a fixed point in the clip, most often the first or last frame, so the model generates the motion between. Google’s documentation calls this generating video “by specifying the first and last frames”.4
What is the difference between a keyframe and a frame?
Every keyframe is a frame, but not every frame is a keyframe. In Blender’s terms a keyframe stores a value you chose, and the frames between keyframes are calculated from them.1
Do all AI video tools let you set keyframes?
Not necessarily. Support depends on the model and the app. Google, for example, lists which of its models accept first and last frames.4 Check your tool’s help pages, and see AI video platforms for an overview.
Learn more on this site
- Start and end images, step by step: first frame and last frame in AI video.
- Animating one picture: image-to-video meaning.
- Making clips from words: text-to-video meaning.
- How the whole film comes together: how AI films are made.
- The words: the AI glossary.
QUICK CHECK · 3 QUESTIONS
Pick an answer to see if you got it. Nothing is recorded.
1. What is a keyframe?
- Any single picture in a video
- A frame where you fix how something looks at a set moment, and the software fills in between
- The last frame a video tool shows before it crashes
Show the answer
Answer: A frame where you fix how something looks at a set moment, and the software fills in between. You set the important moments, like stops on a road trip, and the software or AI works out the frames in between.
2. In AI video, what do keyframes usually mean?
- The start and end images you give the model to move between
- The sound track added after the video is made
- The title text shown at the start of a clip
Show the answer
Answer: The start and end images you give the model to move between. You give a start frame, an end frame, or both, and the model makes a clip that travels between them.
3. How do AI in-between frames differ from classic animation software?
- AI follows a fixed rule, so every run is identical
- AI cannot make in-between frames at all
- AI imagines them, so the path can surprise you and a second run may differ
Show the answer
Answer: AI imagines them, so the path can surprise you and a second run may differ. Animation software calculates the gaps by a fixed rule; an AI model imagines them.
Related terms: Image-to-video · Img2img (image-to-image) · Inpainting and outpainting · Lip sync · LoRA · Negative prompt
- 1Blender, Keyframes: Introduction, Blender Manual, undated, read at source 12 Oct 2026“A Keyframe is simply a marker of time which stores the value of a property.” ↑a“Blender will automatically determine the correct position of the cube for all the frames between frame 1 and 10 depending on the chosen interpolation method.” ↑b“A Keyframe is simply a marker of time which stores the value of a property.” ↑c
- 2Google for Developers, Introducing Veo 3.1 and new creative capabilities in the Gemini API, 15 Oct 2025, read at source 12 Oct 2026“By providing a starting and an ending image, you can direct Veo 3.1 to generate the transition between them”. ↑
- 3Google Cloud, Generate videos using first and last video frames, Gemini Enterprise Agent Platform documentation, last updated 12 Oct 2026, read at source 13 Oct 2026“The following models support generating videos from first and last video frames”: gemini-omni-1.1-flash-preview (preview), veo-3.1-lite-generate-001 (preview), veo-3.1-generate-001 and veo-3.1-fast-generate-001. ↑
- 4Google Cloud, Generate videos using first and last video frames, last updated 10 Oct 2026, read at source 12 Oct 2026“generate videos by specifying the first and last frames of the video”. ↑alists the supported models, two of them marked preview. ↑b
The comparisons and examples on this page (the road-trip stops, the swinging door) are this site’s own ways of explaining, not claims from these sources.
Free printable AI cheat sheets, and AI news every week.
Sixteen one-page AI guides and checklists in one free PDF, no sign-up. And if you want to keep up, the newsletter covers what changed in AI each week, in plain English, with the sources.
SUBSCRIBE FREE ↗ GET THE CHEAT SHEET PACK