Speed Up or Slow Down a Video Without It Looking Broken

When frames have to be invented

If the footage already exists at 30 fps and it needs to run at half speed, the missing frames have to come from somewhere. The crude approach duplicates each frame, which keeps the timing correct and looks exactly as juddery as the arithmetic predicts. The better approach is interpolation: comparing consecutive frames, working out how each part of the picture moved between them, and generating the intermediate image that would have been recorded had the camera been faster.

The frame interpolation tool does this, and it is genuinely impressive on the right material — steady shots, moderate motion, clear subjects. It struggles on the things that break motion estimation everywhere: fast pans, heavy motion blur, objects passing behind other objects, water, fire, and anything where large areas appear or disappear between frames. Those produce warping artefacts around edges that are more distracting than the judder would have been.

The practical rule is to try interpolation on the specific shot rather than trusting it as a policy. It transforms some clips and ruins others, and which category a clip falls into is obvious within two seconds of watching the result.

Speeding up is the easy direction

Going faster just discards frames, and there is always a surplus to discard, so the result is technically clean at any factor. The problems are aesthetic rather than mechanical.

• 1.25× to 1.5× is the tightening range — it removes dead air from a talking-head or tutorial without the viewer registering that anything has been sped up.
• 2× is the point at which speed becomes visible. Fine for a process the viewer is meant to watch impatiently; wrong for anything with speech.
• 4× to 10× is montage territory: assembly, cooking, a walk across a city. Motion becomes comic rather than informative, which is sometimes exactly right.
• 20× — the speed tool's maximum, with Pro (free speed-ups go up to 4×) — is timelapse territory, and at that point the source needs to be long. A ten-second clip at 20× is a half-second flicker.
• Trim first. Speeding up an entire clip to hide a boring middle wastes the good parts too — cut the dead section out and leave the rest at normal speed.

What happens to the audio

Change the speed of a video and the audio has to change with it, or the two drift apart within seconds. There are three ways to handle this and they produce very different results.

Time-stretching is what you want for tutorials, interviews and anything where somebody is talking, and it is what the speed tool applies by default. It holds up well to about 1.5× before the artefacts — a slight metallic edge on sibilants — become audible. Beyond 2× most speech is worth removing and replacing with music. If you are working on the sound separately, the audio speed tool offers the same pitch-preserving stretch on its own.

Speeds that suit each platform

Speed choices are partly about the material and partly about where the video will be watched. Short vertical feeds reward compression of time because the viewer's thumb is the competition; long-form video rewards restraint because the viewer has already committed.

• Reels, Shorts, TikTok. Aggressive. 1.5× to 2× on demonstration, and cut anything that does not move the clip forward. Fifteen seconds of sped-up process beats forty seconds of real time.
• YouTube long-form. 1.1× to 1.25× on the whole edit is common and invisible, and it removes a surprising amount of slack across ten minutes.
• Product and how-to clips. Normal speed for the step that matters, 4× for the waiting. Mixed speed within one clip is more watchable than a single compromise rate (in the speed tool, speed segments within one clip are a Pro feature).
• Sport and action. Half speed on the moment, normal everywhere else — and shoot at 60 fps or higher so the half speed holds up.
• Timelapse of something slow. 20× or more, from footage measured in minutes rather than seconds. 20× is the speed tool's maximum, and a Pro setting.

Order of operations

Speed interacts with everything else you might do to a clip, and doing things in the wrong order costs quality or wastes time. Trim first, so you are not spending processing effort on frames you will delete. Stabilise before changing speed, because stabilisation analyses motion between frames and speed changes corrupt that analysis. Apply speed next. Resize and compress last, once the timeline is final, so the file is only encoded once at the end.

All of this runs on your own device, not our servers, which matters more for speed changes than for most operations: interpolation and time-stretching are computationally heavy, and doing them locally means a long clip starts processing the moment you drop it in instead of after an upload. The free plan takes clips up to 5 minutes and 100 MB, with up to 15 minutes of output; Pro is limited only by your device.