Change Video Speed

How the speed change works

Every frame in a video carries a timestamp saying when it should be shown. Changing speed means rewriting those timestamps: at 2× each one is halved, so a frame that appeared at 10 seconds now appears at 5. This is done with ffmpeg's setpts filter, and because timestamps are part of the compressed stream, the video has to be re-encoded — here to H.264 in an MP4.

The frame rate of the output stays the same as the source. Speeding up therefore drops frames: a 30 fps clip at 2× keeps every second frame, which looks perfectly smooth. Slowing down repeats frames: at 0.5× each frame is shown twice. Nothing is invented between frames — there is no motion interpolation — so slow motion made this way looks stepped rather than silky.

That is why real slow motion starts in the camera. A clip shot at 120 or 240 fps on a phone can be slowed 4× and still have 30 or 60 distinct frames every second. Slow a normal 30 fps clip by 4× and you see each frame for an eighth of a second.

What happens to the sound

Speeding up a recording the old-fashioned way — playing it faster — raises the pitch: chipmunks at 2×, a deep drawl at 0.5×. The audio here goes through ffmpeg's atempo filter instead, which changes tempo while keeping pitch. A lecture at 1.5× sounds like a fast talker, not a cartoon.

atempo works cleanly between 0.5× and 2×, so larger changes are applied as a chain: 4× is two 2× steps and 0.25× is two 0.5× steps. The overall factor is exact, so sound and picture stay in sync to the end.

Quality holds up well for speech from about 0.75× to 2×. At the extremes the time-stretching becomes audible — a faint echo or warble at 0.25×, clipped syllables at 4× — and for very slow motion the sound is often best removed. The Audio option does that.

Choosing a speed

1.25× to 1.5×: lectures, tutorials and talks. Most people follow speech comfortably at 1.5×, and an hour-long recording becomes 40 minutes.

2× to 4×: process videos, walk-throughs, screen recordings with long waits, journeys. A ten-minute drive becomes two and a half minutes at 4× — fast enough to be watchable as a sped-up sequence.

0.5× to 0.75×: following a dance step, a golf swing, a guitar fingering, a sports replay. 0.25×: emphatic slow motion, best on footage shot at high frame rates.

The new length is simple to work out — original length divided by the speed — and the summary shows it before you start: a 3-minute clip is 2 minutes at 1.5× and 12 minutes at 0.25×.

Time and quality

Because the whole video is re-encoded, this is a real encoding job running single-threaded in the tab. Measured on this build, the default setting handles 1080p at roughly 3–4 frames a second on a recent desktop. What matters is the number of output frames: speeding up to 2× halves them and roughly halves the time, while slowing to 0.25× quadruples them.

The Quality setting trades time for size. Balanced (CRF 23, veryfast) is the default; Higher quality is a little sharper and larger; Fastest is about two and a half times quicker and produces a larger file.

If you only need part of the video sped up, trim that part out first with the Video Trimmer, change its speed here, and keep the encoding time to the section you care about.

Changing playback speed versus changing the file

Video players and sites such as YouTube already let you watch at 1.5× or 2×, and if you only want to get through a lecture faster, that is the easier option — nothing needs converting. The speed applies to your own viewing and nowhere else.

Changing the file is for when the speed has to travel with the video: a sped-up clip you are posting to a platform with no speed control, a slow-motion replay for a coach or a team, a time-lapse of a long process for a presentation, or footage for an editor that should arrive already at the right speed.

Private and local

The video is processed by ffmpeg compiled to WebAssembly, inside this browser tab. It is not uploaded, there is no watermark, and the only download is the 32 MB engine on first use.