Why this is fast
Native ffmpeg on a real server is roughly 4–8x faster than browser-WASM ffmpeg for the same job — it has real SIMD, threading, and disk I/O. Uploads land at the nearest Backblaze B2 point of presence via a presigned URL (a viewer in São Paulo is not pushing bytes through US-east), the slow-motion pass runs server-side, and the finished MP4 is delivered through Bunny CDN from the same edge. A typical two-minute screen recording finishes in seconds.
What this tool does, exactly
We push each frame's presentation timestamp through setpts=(1/speed)*PTS and time-stretch the audio with a chained atempo filter (chained because a single atempo only accepts 0.5–2.0, so 0.25x becomes atempo=0.5,atempo=0.5). Video re-encodes with libx264 at CRF 22, audio is AAC at 160 kbps, and the output MP4 carries +faststart for immediate playback. No frame interpolation — we never invent frames the camera did not capture.
When to reach for 0.75x vs 0.25x
0.75x is the gentle setting — it takes the edge off a rushed demo or a fast-talking presenter while still feeling natural to watch. 0.5x is the sweet spot for study and transcription: comfortably slow, audio still perfectly intelligible. 0.25x is for the moments you genuinely need to dissect frame by frame — a UI animation, a hand movement, a sleight-of-hand reveal. The output file gets longer (and a touch larger) the more you slow it, since the same pixels now occupy more seconds, so trim to the segment you care about before exporting if you can.
Sister tools
Need the opposite direction? Video speed controller covers 0.25x to 4x in one slider. Reverse the clip too? Reverse video. Cut down to just the slowed section? Video cutter. Drop the audio entirely for a silent study loop? Mute video. If you would rather let viewers control playback, Clipy is a free screen recorder with a hosted player and built-in speed control.