How to compress video without losing too much quality

Let us be accurate about the premise: compressing a video re-encodes every frame, and re-encoding is lossy. There is no setting that makes a smaller file with no loss at all. What there is, is a large gap between doing it well and doing it badly.

One pass, from the best source

The single biggest avoidable loss is generation loss — encoding something that has already been encoded. Each pass discards detail the previous one left, and the second encoder cannot tell what is already missing. Always start from the original, and if the result is not small enough, go back to the original and use a harder setting rather than compressing your own output again.

Cut pixels before you cut quality

Given a size target, there are two ways to reach it: fewer pixels encoded well, or many pixels encoded badly. The first almost always looks better. Balanced at 720p generally beats Smallest file at 1080p for a similar size, because the encoder has enough bits for the pixels it does have.

Compress last

Trim, crop, rotate and add text first — at the highest quality you have — then compress the finished result once. Doing it the other way means every subsequent edit re-encodes an already-damaged file.

What to expect at each preset

• Higher quality (CRF 22) — very hard to distinguish from the source on normal footage. Use it when the video will be seen large or re-encoded again by a platform.
• Balanced (CRF 27) — the right default. Most people cannot tell at normal viewing size.
• Smallest file (CRF 33) — visible softening in motion and texture. For hard limits only.

Mistakes worth avoiding

• Running the compressor repeatedly on its own output.
• Compressing before editing.
• Compressing hard before uploading somewhere that will re-encode it anyway.