Keyframe Interpolation

In an animation, an object needs to change from frame to frame. In the past, animators drew every single state by hand. Today, one only needs to define so-called keyframes – these are key images that describe a particular state at a particular point in time. The computer calculates everything in between. This automatic filling-in of the gaps is called keyframe interpolation. One example: you set a ball at the top at second 0 and the same ball at the bottom at second 2. The interpolation calculates all the positions in between, producing a smooth fall.

Why keyframes revolutionized animation

Without interpolation, a film running at 24 frames per second would require every single frame to be crafted by hand. A ten-minute scene would then comprise 14,400 individual frames. Keyframe interpolation drastically reduces this effort. The animator only defines the most important moments – the beginning, the end, perhaps a few points in between. The software takes care of the rest.

This not only saves time but also enables precise control. You can move a single keyframe afterward, and all the in-between frames adjust instantly. In the past, the same correction would have meant hundreds of newly drawn frames. This difference is the reason why professional animation software has been built on this principle since the 1980s.

Linear, curve-based, and AI-assisted calculation

The simplest method is linear interpolation: the computer divides the path between two keyframes evenly. If a ball moves from left to right, it covers exactly the same distance in every frame. However, this looks mechanical, because real movements are rarely uniform.

That’s why curve-based methods, so-called splines, are used in practice. With these, the motion can start slowly at the beginning, accelerate in the middle, and slow down again at the end – just like a real object under gravity. The animator controls this curve graphically, similar to bending a line in a drawing program. Modern AI models for video generation extend this principle: they don’t just interpolate position and color, but reconstruct entire scene content between two frames. This makes it possible to generate a smooth video from just a few photos.

A common misconception is confusing interpolation with extrapolation. Interpolation fills gaps between known points. Extrapolation extends a movement beyond the last keyframe into the unknown. Both are useful techniques, but they solve different problems.

Keyframe interpolation in software, film, and AI video

Every time someone sets up a crossfade in a video editing program like Adobe Premiere or CapCut, keyframe interpolation is behind it. In games, too, it ensures that characters move smoothly even though only key poses of the animation are stored. In 3D software like Blender, the so-called Graph Editor is a visible expression of this principle: you can see the motion curves directly and adjust them.

In the current AI debate, the term comes up in connection with video models. Tools like Runway Gen-3 or Google's Veo allow you to specify a starting frame and an ending frame. The model interpolates the entire clip in between. This is technically more complex than classic interpolation, but it follows the same basic principle: define key states, calculate the gaps. The difference is that an AI model doesn’t just account for coordinates, but also textures, light, and motion physics.

For users, this means: what used to keep a team of animators busy can now be triggered with two images and a single click. Keyframe interpolation is not a new concept here – it has simply grown into a new, significantly more powerful form.