Video
Reducing video size: what the compression levels really change, with measurements
A smartphone video quickly weighs several hundred megabytes: too much for an email, slow to send through messaging apps, bulky on a disk. Shrinking it is always a trade-off between size and quality, but that trade-off often stays abstract: what do balanced or smallest size actually produce? To answer, we compressed a real video with each level FileXvert offers, measuring the resulting size and the picture fidelity. The results let you choose with full knowledge.
What makes a video heavy
A video's size is its duration multiplied by its bitrate, that is, the amount of data used for each second. That bitrate depends on three factors: resolution (1080p has 2.25 times as many pixels as 720p), frame rate (a 60 fps video has twice as many frames as a 30 fps one) and the compression level chosen by the encoder.
Content matters enormously too. A static shot of a face compresses very well, because almost nothing changes from one frame to the next. A scene with lots of movement, water, foliage, grain or sand is far more costly: the encoder has to describe changes everywhere. Two videos of the same duration can therefore differ in size by a factor of five.
Our test video and method
We deliberately picked a demanding case: an 89-second smartphone video shot in 1080p at 60 frames per second, in a desert landscape, with continuous camera movement and lots of fine texture (sand, pebbles). The original file, in H.264 with uncompressed sound, weighs 261.3 MB, a video bitrate of about 22 Mbit/s.
Each version was produced with FFmpeg using exactly the parameters FileXvert applies. To measure fidelity, we used the SSIM index, which compares the structure of each frame with the original: 1 means identical. On video, an SSIM around 0.98 remains hard to tell from the original in normal playback; around 0.95, fine details soften on a paused frame; below 0.90, the loss is visible in full screen.
Results: FileXvert's three levels
Here are the results in the most common case, a recent browser on a computer, where FileXvert can use several processor cores.
| Level | Setting applied | Size | Reduction | SSIM |
|---|---|---|---|---|
| Original | — | 261.3 MB | — | 1 |
| No reduction | H.264, default CRF 23 | 181.5 MB | − 31% | 0.977 |
| Balanced | CRF 28 | 77.5 MB | − 70% | 0.955 |
| Smallest size | CRF 32 and down to 720p | 16.8 MB | − 94% | 0.868 |
CRF (constant rate factor) is the quality setting of the H.264 encoder: the higher it is, the stronger the compression. Each 6-point step roughly halves the bitrate, which our figures confirm: from 23 to 28, the size drops from 181.5 to 77.5 MB.
The balanced level offers the best ratio: it removes 70% of the size for a loss that stays discreet, mostly visible on the sand texture in a paused frame. The smallest size level divides the file by more than fifteen, but by dropping to 720p and compressing hard: perfect for a phone or a preview, not enough for a large screen.
Browsers without multi-core support
To use several cores, FFmpeg needs the page to benefit from cross-origin isolation. Some browsers or setups, especially in companies, do not allow it. FileXvert then falls back to a single-core build, and adjusts its settings so that conversion remains bearable: encoding in the fastest mode, and width capped at 1280 pixels. That fast mode compresses much less efficiently, as the same measurement shows:
| Level | Setting applied | Size | Reduction | SSIM |
|---|---|---|---|---|
| No reduction | CRF 23, 720p | 320.8 MB | + 23% | 0.933 |
| Balanced | CRF 28, 720p | 143.4 MB | − 45% | 0.896 |
| Smallest size | CRF 32, 720p | 58.8 MB | − 77% | 0.849 |
A counter-intuitive but important result: in this mode, converting with no reduction can produce a file heavier than the original, even though the resolution went down. If your goal is to make a video lighter and conversion seems slow, always pick a reduction level, or try again from a recent browser on a computer.
Which level should you choose?
Start from the final use rather than from a number:
- Archiving, editing or showing on a large screen: do not compress, or use no reduction only to change format.
- Sending by email, publishing on a website, sharing in a work conversation: balanced, which divides the size by three without any bothersome flaw.
- A strict upload limit, or a video meant to be watched on a phone: smallest size.
- A video several minutes long to send by email: no level will be enough; use a sharing link.
One last tip: always start from the original. Compressing an already compressed video accumulates losses; if the result does not suit you, start again from the source with another level rather than recompressing the result.
Checking the result before sending
Watch the converted video at the beginning, middle and end, ideally on the screen where it will be seen. Compression flaws show up first in dark scenes, as blocks, in fast motion, which turns blurry, and in fine text, which smears. Also check that sound and picture stay in sync on a spoken passage.
If the result is not good enough, start again from the original with a less aggressive level. The free plan handles files up to 100 MB; a longer video, like the one in this measurement, requires the Premium plan or trimming beforehand.