First, know why the file is big
File size is bitrate × duration — nothing else. So a file is large for exactly one of two reasons: it is long, or a lot of data is being spent per second. That second case has its own usual suspects: high resolution (4K carries four times the pixels of 1080p), high frame rate (60 fps footage from phones and game captures), a generous or barely-compressed source (screen recorders and phone cameras err heavily on the side of quality), or an old, inefficient codec. Checking the file's duration, resolution, and rough size-per-minute tells you which lever will pay off most before you touch anything.
The levers, in the order to pull them
1. Trim what nobody needs to see
Cutting footage is the only step that is completely free: no quality is lost on what remains, and the savings are exactly proportional. Thirty seconds of highlight sent instead of a five-minute recording is a 90% reduction before any compression enters the picture. It sounds too obvious to list — and it is the step people most often skip on their way to fiddling with encoder settings.
2. Step the resolution down (when it's overkill)
Resolution is the heaviest single lever because pixel count grows with the square of the dimensions: 4K → 1080p is a 75% cut in pixels, 1080p → 720p roughly halves them again. The savings in bytes are nearly as dramatic. The question to ask is where the video will be watched: a clip destined for a chat app or a web page embed will be viewed on a small area of a phone screen, where 720p is indistinguishable from 4K. Keep original resolution only when viewers will genuinely watch full-screen on large displays.
3. Raise the compression (CRF)
The quality setting is the fine-grained lever. Modern encoders use a constant-quality scale called CRF where a few points make a big difference: roughly, +6 CRF ≈ half the size. Moving from a visually-lossless 18 to the standard 23 typically halves the file with no visible change on normal viewing; 23 → 28 halves it again with only mild softening. Diminishing returns arrive quickly past that, as compression artifacts become visible. The full scale and how to pick a value are covered in our CRF guide.
4. Don't forget the audio
Audio is a minority of most files, but not a negligible one. Uncompressed PCM audio (common in screen recordings and camera footage) costs about 1.5 Mbps — some 11 MB per minute — where 128–192 kbps AAC sounds identical for typical content at a tenth of the size. And if the video track is not needed at all — a recorded talk you only listen to — extracting the audio shrinks the file by 90% or more in one move. Which audio format to extract to is its own small decision, covered in our audio formats guide.
5. Let a modern codec help
Codec efficiency compounds with everything above. Files from older sources (AVI-era codecs, motion-JPEG screen captures) shrink dramatically when re-encoded to H.264 at the same visual quality. Going further, VP9 or H.265 saves another 30–50% over H.264 — with the caveat that encoding is slower and older devices may not play the result. For files that must play absolutely everywhere, well-tuned H.264 remains the pragmatic choice; the trade-offs are mapped in our format guide.
What the steps stack up to
| Action | Typical saving | Quality cost |
|---|---|---|
| Trim to the needed section | Proportional to cut | None |
| 4K → 1080p | 60–75% | Invisible on phones/laptops |
| CRF 18 → 23 | ~50% | Invisible in normal viewing |
| CRF 23 → 28 | ~50% | Mild softening |
| 60 → 30 fps | 20–40% | Fine unless sport/gaming |
| PCM → AAC audio | ~10 MB/min | Inaudible for typical content |
The two mistakes that can't be undone
Re-encoding in generations. Every lossy encode discards information, and the losses compound. Compressing a file, deciding it is still too big, and compressing the output again bakes in two rounds of artifacts where one stronger setting would have caused one. Always return to the best original you have and convert once with final settings. Keep that original until you are sure you are done with it.
Expecting compression to be reversible. A file compressed to CRF 30 does not contain its former detail, and no future conversion can bring it back — converting it to a larger, higher-quality format only wastes space. When in doubt, err one step gentler than you think you need: storage is cheaper than re-shooting.
Fast workflow: trim a representative 20–30 seconds, convert it with your candidate settings, and check it on the screen where the real video will be watched. Iterate on the short clip, then apply the winning settings to the full video. Two minutes of testing routinely saves an hour of re-encoding.
A worked example
A 10-minute 4K/60fps screen recording weighing 3 GB needs to land under 100 MB for sharing. Trim dead air at both ends: 8 minutes remain (~2.4 GB). It will be watched in a browser window: 1080p brings it near 700 MB. Motion is mostly a cursor, so 30 fps costs nothing perceptible (~500 MB). CRF 23 → 26 for the final squeeze: about 90 MB, still crisp for screen content. Every step used a lever this guide covered, none of them exotic — and the original stays safely on disk in case a higher-quality copy is ever needed.