Video Speed

Speed a video up or slow it down, from a quarter speed to four times faster, and save it as MP4 or WebM. Keep the voice at its natural pitch or let it rise and fall like a tape, and copy the picture with every frame kept when you want no quality loss. Nothing is uploaded.

Tool Media & Files Updated Oct 16, 2026
How to Use
  1. Drop a video: MP4, MOV, MKV, WebM, AVI and most other formats. The first time, the page loads a video engine of about 32 MB.
  2. Set the speed from 0.25× to 4×. The player previews it and the panel shows the new length and frame count.
  3. Choose the picture mode: re-encode at the original frame rate, or copy every frame with no quality loss (the frame rate changes instead).
  4. Choose the sound: natural pitch, pitch that moves with the speed, or no sound; and, when re-encoding, MP4 or WebM.
  5. Press Make video. The result is read back for its real length and frame count; download it when it looks right.
Input
Drop a video, or
MP4, MOV, MKV, WebM, AVI — processed on your device, never uploaded
×, 0.25–4
Presets
Old and New Timeline
New length
—
Speed
—
Frames
—
File size
—

Worked Example

The built-in test clip at 1.5×. The clip is 4.00 seconds of a moving test pattern at 30 fps, 120 frames, with a 440 Hz tone at 48,000 Hz. At 1.5× it lasts 4 ÷ 1.5 = 2.67 seconds. Re-encoded at 30 fps that is 2.667 × 30 = 80 frames, so 40 of the 120 are dropped. The sound goes through atempo=1.5, so the tone is still 440 Hz, just over sooner.

Pitch and copy modes. With “pitch moves with speed” the 48,000 samples of each second are labelled as 72,000 per second and resampled back to 48,000, so the 440 Hz tone becomes 660 Hz, 12 × log₂ 1.5 = 7.02 semitones higher. In copy mode at 2× nothing is dropped: all 120 frames are kept, their timestamps halved, and the clip plays at 60 fps for 2.00 seconds without the picture being re-encoded.

The common mistake: expecting slow motion to add frames. At 0.25× the 4-second clip lasts 16 seconds, but it still contains only 120 different pictures. Re-encoded at 30 fps each picture is shown 4 times (480 frames); copied, the clip plays at 7.5 fps. Either way it looks jerky. Smooth slow motion needs footage shot at a high frame rate: 120 fps at 0.25× still gives 30 different pictures per second.

Show Work

Drop a video or press a preset to see the new length, frames and filters, then the result read back.

Formulas

New length
T′ = T ÷ speed
10:00 at 2× is 5:00; at 0.5× it is 20:00
Frames, re-encoded
n′ = T′ × original fps
Faster drops frames, slower repeats them
Frame rate, copied
fps′ = fps × speed
Every frame kept; 30 fps at 2× plays at 60 fps
Pitch change
semitones = 12 × log₂(speed)
Only in “pitch moves” mode; 2× is +12, 0.5× is −12
atempo chain
speed = t₁ × t₂ × …, each 0.5–2
4× is 2 × 2; 0.25× is 0.5 × 0.5

Timestamps, Tempo and Pitch

A video file does not store a speed. Each frame carries a presentation timestamp saying when to show it, so changing speed means changing timestamps: ffmpeg’s setpts filter divides them by the speed, and its -itsscale option multiplies the timestamps of a whole input before anything is decoded, which is why the picture can then be copied untouched. Sound is different because its timing is the sample rate itself.

Playing samples faster raises the pitch, which is what happens with a tape or a record at the wrong speed; ffmpeg’s asetrate filter does exactly this, “a change of speed and pitch” in its own words. Changing tempo while keeping pitch needs time-stretching: atempo, which its source code describes as an implementation of the WSOLA algorithm, cuts the sound into short overlapping pieces and lays them closer together or further apart. Its documentation says tempos above 2 skip some samples rather than blend them, and recommends chaining several atempo instances instead, so this tool always keeps each stage between 0.5 and 2.

About This Tool

This tool makes a real new video at a different speed, from 0.25× to 4×. It shows the new length and frame count before you encode, previews the speed in the player, and afterwards reads the new file back with ffmpeg to report its real length, streams and number of frames. It can copy the picture with every frame kept, which loses no quality, or re-encode it at the original frame rate for the widest compatibility.

Everything runs in your browser on a WebAssembly build of ffmpeg, the same engine as the Video Converter, and your video is never uploaded. Without a video, the presets use a 4-second test clip that the engine makes on your device.

It is for students speeding through lectures, creators making timelapses and slow-motion clips, and anyone fitting a clip into a time limit.

Related tools: Video Converter, GIF from Video, and Video Inspector.

Frequently Asked Questions

What is re-encoded and what is copied?

The sound is always re-encoded, as AAC (or Vorbis in WebM), because its samples have to change. The picture is re-encoded in the default mode, as H.264 at CRF 22 or VP8, which costs one generation of quality. In “Copy every frame” mode the picture is not re-encoded at all: ffmpeg rescales each frame’s timestamp and copies the compressed frames, so it is lossless and much faster.

Does the voice sound like a chipmunk?

Not by default. The sound is time-stretched with ffmpeg’s atempo filter, which keeps the pitch, so speech at 1.5× sounds faster but not higher. If you choose “pitch moves with the speed”, the samples are simply played at a different rate, like a tape: 1.5× raises the pitch by 12 × log₂ 1.5 = 7.02 semitones and 2× by a full octave.

Why is slow motion jerky?

Slowing down cannot invent frames that were never filmed. Re-encoding at the original frame rate repeats frames (each one 4 times at 0.25×), and copy mode keeps every frame but shows them less often, so a 30 fps clip at 0.25× plays at 7.5 fps. Footage shot at 120 or 240 fps gives smooth slow motion.

Which picture mode should I use?

Copy is best for speeding up when you want perfect quality and a fast result: 2× of a 30 fps video becomes 60 fps with every frame kept. Re-encode when you need a normal frame rate for an upload, when you are slowing down, or when copy mode cannot write your video’s codec.

How long does it take, and is my video uploaded?

Nothing is uploaded: ffmpeg runs on your device in a WebAssembly build. Re-encoding uses one CPU core and is often slower than real time for 1080p; copy mode takes seconds. The whole file is held in memory, so the practical limit is about 1 GB, less on a phone.

How do I use the Video Speed?

Simply type your numbers and read the result, which refreshes the instant you change something. There is nothing to submit and nothing to wait for.

Do I need to install or sign up for anything?

Not at all — it runs in the browser with nothing to install and no account. After it loads once, it even works without an internet connection.

Is my information private?

Yes. Everything happens in your browser. Nothing you type is sent to a server or saved anywhere.

Common Use Cases

Lectures and tutorials at 1.25×

A 60-minute lecture becomes 48 minutes with the voice still at its natural pitch.

Timelapse from a screen recording

At 4× a 20-minute build becomes 5 minutes; at 30 fps three of every four frames are dropped.

Slow motion of a swing or a jump

A 2-second clip at 0.25× lasts 8 seconds; shot at 120 fps it still plays at a smooth 30 fps.

Fit a time limit

A 72-second clip at 1.2× is exactly 60 seconds for a platform with a one-minute limit.

Last updated: