Compress Video

Two ways to compress, and when each is right

Compression works by throwing away detail the eye is least likely to miss, and the only real question is how much. There are two ways to express the answer, and they suit different problems.

A quality level uses constant-quality encoding: the encoder spends whatever bitrate each moment needs to hit a consistent visual standard, so a still shot of a wall costs almost nothing and a crowd scene costs a lot. The result looks even throughout, but you cannot know the file size in advance. Use this when you simply want the video smaller and have no hard limit to clear.

A target file size works backwards from the budget. The bitrate is calculated from the clip's actual length, the audio track's share is subtracted, the encode runs, and then the output is measured — and re-encoded once with a corrected bitrate if it overshot. Use this whenever something must clear a specific limit, which is most of the time.

The limits people are usually trying to clear

It is worth setting the target slightly under the stated cap rather than exactly at it, since some services count the whole upload rather than just the file.

• WhatsApp — 16 MB for a video sent in a chat (larger files re-compressed by WhatsApp, or sent as a document up to 2 GB)
• Email (Gmail, Outlook) — 25 MB per message, attachments included
• Discord — 10 to 20 MB without Nitro (Discord has been changing this; check the limit it shows you)
• Most job portals and course uploads — usually between 20 MB and 100 MB

When the result still looks poor

A hard size target on a long clip is a hard constraint, and there is no encoder setting that escapes arithmetic: ten minutes squeezed into 8 MB has roughly 100 kilobits per second to work with, and no amount of cleverness makes that look good at 1080p.

The productive move in that situation is to reduce what the encoder has to describe rather than to demand it describe the same thing in fewer bits. Shrinking the frame is by far the most effective: the same 8 MB spread over 720p instead of 1080p gives each pixel more than twice the data, and a smaller, sharper picture reads as better quality than a full-size, mushy one. Trimming the clip shorter has the same effect. Removing the sound frees up around a megabyte per minute.

Why compressing sometimes makes a file larger

It is a genuinely counter-intuitive result, and the tool will tell you when it has happened rather than letting you discover it. Re-encoding a video that is already efficiently compressed — one that has been through a social platform, say, or was exported at a sensible bitrate to begin with — can add bytes rather than remove them. The encoder is being asked to faithfully reproduce the compression artefacts the previous encoder introduced, and describing noise is expensive.

The answer is not to compress more gently but to compress harder: choose a lower quality level, reduce the frame size, or switch to an explicit target size, which forces the bitrate down regardless of what the source looked like.