ConvertBox Logo
ConvertBox

How to reduce video file size: bitrate, CRF and resolution

A video's size is its bitrate multiplied by its length, so there are only three ways to shrink it: make it shorter, spend fewer bits per second, or use a more efficient codec. Trimming and downscaling usually save more than pushing the quality setting.

ConvertBoxPublished

File size is bitrate multiplied by duration

A video file's size follows from one piece of arithmetic. Bitrate is the number of bits spent per second of playback, for video and audio together. Size in megabytes equals the bitrate in megabits per second, times the length in seconds, divided by 8. A 60-second clip with 8 Mbps video and 128 kbps audio is 8.128 Γ— 60 Γ· 8, about 61 MB.

Resolution and frame rate do not appear in that formula. They matter indirectly: more pixels and more frames need more bits to look clean, so the encoder is given, or chooses, a higher bitrate. This is why two 1080p files of the same length can differ in size by a factor of ten, and why there are only three ways to shrink a video: make it shorter, give it fewer bits per second, or make those bits go further with a more efficient codec.

What a minute of video weighs at common bitrates

To find a file's bitrate, multiply its size in megabytes by 8 and divide by the duration in seconds. The sizes in this section are plain arithmetic; the typical uses are rough guides for H.264, since the bitrate a scene needs depends heavily on how much it moves.

Total bitrateSize per minuteRoughly typical of
1 Mbps7.5 MBScreen recordings, slides, 480p video
2.5 MbpsAbout 19 MB720p video for the web
5 Mbps37.5 MB1080p video for the web
8 Mbps60 MBHigh-quality 1080p
16 Mbps120 MB1080p straight from a phone camera
45 MbpsAbout 340 MB4K straight from a phone camera

CRF: asking for a quality level, not a size

Encoders can be driven in two ways. With a target bitrate you fix the size and let quality vary. With a constant rate factor (CRF) you fix the quality and let the bitrate vary: a static slide costs almost nothing, handheld footage of moving water costs a great deal. CRF gives better-looking results per megabyte, but the output size cannot be known before encoding.

For the x264 encoder the scale runs from 0 to 51, lower meaning better quality and larger files, with 23 as the default. A widely used rule of thumb is that adding 6 roughly halves the bitrate. If a clip comes out at 40 MB at CRF 23, expect about 20 MB at 29 and about 10 MB at 35, by which point blocking in motion and in dark areas is visible.

ConvertBox's Compress Video tool works this way: a CRF slider from 18 to 40 starting at 23, x264 at the medium preset, and audio re-encoded to AAC at 128 kbps. It has no target-size mode, so reaching a specific size takes trial and error.

Resolution and frame rate: the larger levers

A 4K frame (3840 Γ— 2160) holds 8.3 million pixels, a 1080p frame 2.1 million, a 720p frame 0.9 million. Going from 4K to 1080p removes three quarters of the pixels; from 1080p to 720p, more than half. The bitrate needed does not fall quite in proportion, because larger frames compress somewhat more efficiently per pixel, but downscaling is still the biggest single saving available for high-resolution footage.

The trade-off differs from raising the CRF. A 1080p file squeezed to 2 Mbps shows blocks and smearing; the same clip downscaled to 720p at 2 Mbps is softer but clean, and on a phone screen the lost resolution is hard to see. As a rule, once artefacts become visible at the size you need, drop the resolution instead of pushing the quality setting further. Halving the frame rate from 60 to 30 fps also helps, though by less than half, since the remaining frames differ more from each other.

The Compress Video tool never changes resolution or frame rate; a 4K clip stays 4K. To downscale, use a desktop encoder such as HandBrake or FFmpeg, or export a smaller version from the phone or editing app that produced the file.

Two cheap savings: trim first, then check the audio

Because size is proportional to duration, cutting a 90-second clip down to the 30 seconds that matter removes two thirds of the file before quality is touched. The Trim Video tool sets its cut points in whole seconds and re-encodes the selected range with H.264, so the size of the trimmed file is set by the encoder, not by the original bitrate.

Audio is usually a small share. At 128 kbps it adds 0.96 MB per minute: under 2% of an 8 Mbps video, but about 20% of a file squeezed to 500 kbps of video, where a lower audio bitrate starts to matter. And when only the sound is needed, as with a lecture or a meeting, drop the picture altogether: extracted as a 192 kbps MP3, an hour is about 86 MB however large the video was.

Fitting under an upload or attachment limit

Limits differ between services and change over time, so look up the current figure for the one you are using, then work backwards from it.

  • Compute the bitrate budget: limit in MB Γ— 8 Γ· duration in seconds. For a 25 MB limit and a 2-minute clip that is 200 Γ· 120, about 1.67 Mbps in total, or roughly 1.5 Mbps for the picture after 128 kbps of audio. That is workable at 720p and too thin for clean 1080p.
  • Leave headroom for email. Attachments are Base64-encoded in transit, which adds about a third to their size, so a limit that counts the encoded message, as mail servers commonly do, leaves room for a file of roughly 18 MB out of 25.
  • Expect messaging apps to re-compress video on their own. If the app offers sending as a file or document instead of as media, the video usually arrives unaltered, subject to that option's own limit.
  • If the budget comes out below about 1 Mbps for footage with real motion, stop compressing and send a link from cloud storage instead.

When re-encoding does not help, and the ceiling of browser encoding

Compression only saves space when the source was wasteful. HEVC video from a recent phone, or a file a streaming site has already squeezed, can come out larger when re-encoded to H.264 at a low CRF; the Compress Video tool then reports a reduction of 0, and the fix is a higher CRF. Its output is always H.264 in an MP4 container even though the download keeps the source extension, so if an app refuses the result, rename it to .mp4.

Encoding in a browser tab runs on WebAssembly, far slower than a native encoder, and both the source and the result are held in memory. The compressor's picker accepts up to 500 MB, but phones and low-memory laptops may fail well below that. The Convert Video tool is stricter still, with a 50 MB cap and a fixed CRF of 28. For anything long or 4K, or needing HEVC, hardware acceleration, downscaling or an exact target size, a desktop encoder is the right tool.

Tools mentioned in this guide

Related guides

All guides