What Is Motion Interpolation?
Motion interpolation is a technique for increasing a video's frame rate by generating new frames that show moving objects at positions between where they appear in the recorded frames. It is the technology behind the "motion smoothing" settings on many televisions, and it is how tools like SmoothMotion.video turn 30 FPS footage into 60 or 120 FPS. This article explains how it works, step by step, why it sometimes goes wrong, and when it is worth using.
The problem it solves
Video is a series of still images. When a ball moves across the screen in a 30 FPS clip, you see it in 30 distinct positions each second. If you want 60 FPS, you need 30 more positions – but the camera never recorded them.
The naive approaches are unsatisfying. Repeating frames raises the frame count without adding any new information, so motion looks the same. Blending frames (averaging two neighbours) produces a frame containing two faint copies of the ball rather than one ball in the middle. Motion interpolation instead tries to work out where the ball went and draw it halfway along that path.
How motion interpolation works
Most classical motion interpolation algorithms follow the same broad pipeline.
1. Divide the frame into blocks
The image is split into a grid of small squares, called macroblocks, typically 8×8 or 16×16 pixels. Working in blocks rather than individual pixels keeps the computation manageable and makes matches more reliable, since a block contains enough texture to be recognisable.
2. Motion estimation
For each block in one frame, the algorithm searches the neighbouring frame for the area that looks most similar. Similarity is usually measured by summing the differences between pixel values; the lower the total, the better the match. Searching every possible position (an exhaustive search) gives the best answer but is slow, so practical algorithms use smarter search patterns – such as diamond, hexagon or predictive zonal searches – that test promising positions first and home in on the best one.
Motion estimation can be done in one direction (from the previous frame to the next) or bidirectionally, comparing in both directions. Bidirectional estimation is more robust, because a match that looks plausible one way but not the other can be treated with suspicion.
3. Motion vectors
The result of motion estimation is a motion vector for each block: an arrow describing how far, and in which direction, that block moved between the two frames. A field of motion vectors is effectively a map of movement across the whole picture. Good algorithms also smooth this field, because neighbouring blocks on the same object usually move in similar ways, and an outlier vector is often a mistake.
4. Motion compensation
To create a frame at a point between the originals – say exactly halfway – each block is moved part of the way along its motion vector. A block that moved 20 pixels to the right is drawn 10 pixels to the right in the halfway frame. Pixels are typically taken from both neighbouring frames and combined, which reduces noise and helps fill gaps.
Moving blocks independently can leave visible seams where blocks meet. Overlapped block motion compensation reduces this by letting blocks overlap slightly and blending them with a weighting window, so edges fade into one another. Variable-size block motion compensation goes further, splitting blocks into smaller ones where motion is complex, such as around the edge of a moving object.
5. Scene change detection
If the video cuts from one shot to another, there is no meaningful motion between the last frame of one scene and the first of the next. Good implementations detect this – often by measuring how different the two frames are overall – and fall back to a simpler method for that transition rather than trying to morph one scene into the other.
If you want to see how these stages fit into a real processing pipeline, our page on how SmoothMotion.video works follows a video from upload to download.
A short history: motion interpolation in TVs
Motion interpolation became familiar to the public through television. As flat-panel TVs grew larger and moved to 100/120 Hz and faster panels, manufacturers added processing that interpolated broadcast and film content up to the panel's refresh rate. Their aims were to reduce the blur that LCD panels showed on fast motion and to smooth out the judder of 24 FPS film on screens refreshing at other rates.
Each manufacturer branded the feature differently, and it was often switched on by default. That led to a widespread reaction against it, particularly among film enthusiasts and filmmakers, who argued that it changed the intended look of films. Many TVs now include a "filmmaker" or similar mode that switches interpolation off.
The soap-opera effect
The most famous side effect of motion interpolation is the soap-opera effect. Cinema has been shot at 24 FPS for roughly a century, and audiences associate its slight motion cadence with "film". Television soap operas and live broadcasts, by contrast, were traditionally shot on video at higher temporal rates. When film is interpolated to 60 FPS or higher, it loses the 24 FPS cadence and starts to look like video – which many viewers perceive as cheap or overly real.
Whether this is a problem depends entirely on the content and the viewer. For sports, gameplay, screen recordings, home videos and drone footage, extra smoothness is usually welcome. For films and dramas, many people prefer the original frame rate. The 30 FPS vs 60 FPS comparison covers this trade-off further.
Common artifacts and why they happen
Motion interpolation is an educated guess, and some situations make guessing hard:
| Artifact | Cause |
|---|---|
| Warping or rubbery edges | Motion too fast for the search range, or motion vectors that are wrong |
| Ghosting or halos around objects | Occlusion: background revealed or covered as objects move, with no single correct source to copy |
| Garbled frame at a cut | Scene change not detected, so two unrelated images are morphed |
| Shimmering on fences, grilles or stripes | Repeated patterns produce several equally good matches |
| Wobbling text or logos | Static overlays sit on top of a moving background, and blocks contain both |
| Uneven movement in animation | Held frames (drawn "on twos" or "on threes") mean there is no motion to interpolate, then a sudden jump |
| Breakup on very fast motion | Large displacements fall outside the search area |
Interpolation also cannot add real detail. It repositions what is already in the frames; it does not recover information the camera never captured, sharpen blur or increase resolution.
Why block-based interpolation is predictable
Block-based motion interpolation is fully transparent: every stage described above – block division, motion search, vector smoothing, compensation and scene change detection – is a known, well-documented technique with tunable parameters. The same input always produces the same output, and when artifacts appear they can be traced back to a clear cause, such as an occlusion or a missed scene cut. That consistency is why SmoothMotion.video uses bidirectional motion search with overlapped block motion compensation: you can try a short clip and expect the full video to behave in the same way.
Try it yourself
The best way to understand motion interpolation is to see it on your own footage. Upload a clip to SmoothMotion.video and compare the result side by side with the original. For more background, read our introduction to video frame interpolation, or our practical guide to smoothing choppy video.