Speed Up Video

Speed Up Video re-times a clip with ffmpeg's setpts filter on the OmniKitly processing service. Six speeds are offered — 0.25×, 0.5×, 1.5×, 2×, 4× and 8× — and the new running time is the old one divided by that figure, so a ten-minute recording comes back at two and a half minutes on 4× and at twenty minutes on 0.5×. What setpts moves is the moment each frame appears; on its own it neither manufactures a frame nor keeps one it no longer has room for. Speeding up therefore drops what will not fit at the source's rate: a four-second 30fps test clip of 120 frames came back as 32 frames, still 30fps, at 4×. Slowing down does the opposite and keeps all 120, lowering the rate instead — the same clip at 0.25× returned those 120 frames spread over sixteen seconds at 7.5fps, and at 0.5× it returned them at 15fps.

That has two consequences worth expecting. A slowed clip holds each original frame on screen for longer — four times as long at 0.25× — so quick movement steps visibly rather than flowing, and the low output frame rate can look wrong to an editor or a platform that expects 24fps or more; the blending option below is the answer to both. Speeding up is the mirror image: an 8× pass keeps roughly one frame in eight and discards the rest, so the file shrinks broadly in step with the duration. That sampling is the same thing a timelapse does, which is why this tool gives the timelapse look — what it will not do is hold on to the frames it skipped.

The audio setting offers three routes, and one of them does not currently do what its label says. Keep the pitch sends the track through ffmpeg's atempo filter, which stretches or compresses the waveform in time and leaves the frequencies alone, so speech at 2× is hurried but still recognisably the same person; atempo handles a factor of only 0.5 to 2.0 in a single stage, so the extremes are built by chaining stages and 8× is three doublings in a row. Drop the audio removes the track before encoding. The third, Let the pitch change, is the one to leave alone for now: it rewrites the audio timestamps and never touches the samples, so the exported track still carries every original sample at its original rate — the pitch does not rise, the sound does not shorten, and a four-second clip run at 4× ends up with four seconds of audio against one second of picture. Use Keep the pitch or Drop the audio until that is put right. On either of those the sound is re-encoded to AAC rather than copied across.

Two fixed limits and one option to know before you start. The result is always MP4 carrying H.264 and AAC at one fixed constant-quality level, so an MKV, WebM or MOV source changes container as well as tempo, and neither the format nor the bitrate is yours to set here. The speed applies to the whole clip, with no ramp between two speeds and no way to treat one section differently. The option is the blend-frames checkbox, which appears only at 0.25× and 0.5× and stays off until you tick it: with it on, ffmpeg's minterpolate filter fills the stretched gaps with cross-faded frames at up to thirty a second, so an edge dissolves towards its next position instead of jumping to it. That multiplies the frames the encoder has to write, so the service works out whether the clip fits before it starts and refuses the combination with the size it can manage rather than running out of time halfway through.

Secure server processing. This operation requires temporary server processing.