How to compress a video without ruining it
Compression is a negotiation between size and detail, and most people lose it by pulling the wrong lever. Understanding what the encoder is actually doing makes the difference between a clean small file and a blocky one.
Updated 2026-07-14
What compression actually throws away
A video codec does not store pictures. It stores a handful of complete frames, called keyframes, and then describes everything that follows as the changes from the frame before. A static shot of a person talking barely changes between frames, which is why it compresses beautifully. A handheld shot of leaves moving in wind changes everywhere at once, which is why it compresses terribly.
Within each frame, the encoder discards detail that human vision is bad at noticing, keeping the edges and shapes your eye tracks and being far more approximate about fine texture and subtle gradients. Lower the bitrate and you are simply telling it to be more aggressive about that, and past a certain point the approximations become visible: blocking in dark areas, banding in skies, and a smeared look on anything moving fast.
Target a size, not a quality slider
A quality slider asks you to guess. A target size does not, because the arithmetic runs the other way and the encoder can solve for it: take your size budget, subtract what the audio needs, divide by the duration, and that is your video bitrate.
That is why compressing to fit Discord's 10 MB or Gmail's 25 MB should never be trial and error. You know the number. The tool can work backwards from it, and then verify that the file it produced actually lands under the cap rather than hoping.
It also means the tool can tell you when what you are asking for is impossible. A 30 minute video will not become a good 10 MB file, because the bitrate that implies is far below what any codec needs to produce a watchable picture. An honest compressor says so instead of handing you a smear.
Lower the resolution before you starve the bitrate
This is the single most useful thing to know about compression, and it is the step almost everyone skips. Resolution and bitrate are not independent knobs, they are a matched pair, and a bitrate that looks terrible at one resolution can look clean at another.
The reason is straightforward. A 1080p frame contains about 2 million pixels, and describing all of them takes data. Starve it and the encoder has no choice but to approximate every one of them badly, which is what blocking is. The same bitrate spent on a 720p frame, which has less than half as many pixels, gives each one more than twice the data, and it looks noticeably better.
So when a target size is tight, step the resolution down first. A clean 720p is a far better video than a mushy 1080p, and on a phone screen, which is where most video is watched, almost nobody can tell the difference to begin with.
- 1080p wants roughly 4 to 8 Mbps to look genuinely good.
- 720p is comfortable from about 2 to 4 Mbps.
- 480p holds up around 1 Mbps, which is where you land when a target is really tight.
- Below roughly 1 Mbps, drop the resolution again rather than pushing the bitrate lower.
The free wins to take first
Before trading away any quality at all, take the reductions that cost you nothing. They are frequently enough on their own.
- Trim the dead air. Size scales directly with duration, and the frames you keep are completely untouched. This is the only truly free compression there is.
- Remove the audio if nobody is speaking. A silent audio stream still carries a bitrate for the whole video.
- Do not re-compress an already compressed file if you still have the original. Every lossy generation compounds the last one's damage, and the second encode faithfully preserves the first encode's artifacts.