A smaller video is not necessarily the same quality. Keep the original, compress a copy, then watch that copy before sharing it. Start with the quality mode in the video compressor unless you have a strict upload limit.
Three real encodes of the same clip
On September 10, 2026, we encoded one 10-second, 1080p, 30 fps animation with no audio. These are native FFmpeg 8.1 measurements on Windows, using H.264 and the fast preset, one run per setting. The source is 30,704,510 bytes. MB below means 1,000,000 bytes.
| Setting | Output bytes (MB) | Reduction |
|---|---|---|
| CRF 23 | 7,246,823 (7.25 MB) | 76.4% |
| CRF 28 | 3,337,442 (3.34 MB) | 89.1% |
| CRF 36 | 1,035,620 (1.04 MB) | 96.6% |
A larger CRF generally trades more detail for a smaller file. These reductions apply to this clip, not to every phone recording. An already compressed source may have much less room to shrink.
Compare the same frame at three seconds
Each image was decoded from the actual output and resized to 960 pixels wide. Inspect the grass, leaf edges and shaded areas. A scaled still image does not fully reveal motion artifacts or full-resolution detail.
Source
CRF 23: 7.25 MB
CRF 28: 3.34 MB
CRF 36: 1.04 MB
The compressor’s comparison slider lets you move the boundary between the source and the encoded frame.
What to check in your own video
- Read small text. Look at subtitles, game UI and numbers in presentations at the size your recipient will use.
- Watch movement. Camera pans, water and fine textures can show artifacts that a single still frame misses.
- Check sound and the ending. Play the start and end to catch missing audio, wrong duration or an incomplete export.
If the copy looks worse than you can accept, return to the original and use a gentler setting. Recompressing the degraded copy cannot restore lost detail.
Choose the right way to make sharing easier
If only one moment matters, trim the clip first. If the whole recording matters, try quality-based compression. If preserving the original file matters more than sending an attachment, share the original through your existing cloud storage and check recipient permissions.
Lowering resolution may reduce size but can make small text harder to read. There is no universal resolution that looks good for every use case.
For a size-based sharing copy, the WhatsApp compressor and Slack compressor compare a custom target with actual output bytes. Presets are not universal service limits. If the details your recipient needs become unreadable, consider a sharing link or a shorter excerpt before increasing compression.
What this test does not establish
All three outputs decoded successfully from start to finish and contained 10 seconds of 1920×1080 H.264, yuv420p video. This run did not test audio preservation, live-action footage, HDR, phones, subjective quality or VMAF. It does not measure the speed of this website’s WebAssembly tools.
Raw measurements, commands and SHA-256 hashes are available. Reproduce the run with scripts/measure-sharing.ps1 in the repository. Recorded times are single observations including process startup, not averages.
Footage and frames: Big Buck Bunny © Blender Foundation, CC BY 3.0. The source clip was encoded and resized for these comparisons.