Video Compressor

About Video Compressor

This compressor runs ffmpeg.wasm — FFmpeg compiled to WebAssembly, version 0.12.10 with the single-threaded core — inside your browser, downloading it from a CDN the first time you use it. Your video is never uploaded anywhere, but encoding speed is limited to one CPU core. Expect it to be several times slower than desktop FFmpeg; a one-minute 1080p clip can take a few minutes.

There are three modes. Quality compression re-encodes to H.264 (libx264) in an MP4. Keep original (no re-encode) copies the streams into a new container without re-encoding. Remove audio leaves the picture untouched and drops the sound track. Only the first one makes a file substantially smaller.

CRF targets quality, not file size

Quality mode does not set a bitrate. It uses CRF (Constant Rate Factor), a quality target from 0 to 51 where lower means better and bigger. The encoder spends more bits on busy scenes and fewer on static ones to keep perceived quality roughly even, so at the same CRF a football match comes out large and a slideshow recording comes out small.

The FFmpeg wiki's rule of thumb is that raising CRF by 6 roughly halves the file, and lowering it by 6 roughly doubles it. The tool's three steps are 23 (high quality), 28 (default) and 33 (strong compression), five apart. x264's own default is 23, so treat 23 as hard to tell from the source and 33 as fine on a phone screen.

The catch with CRF is that you cannot choose the output size in advance. If a platform caps uploads at a certain size, compress once, and if it is still over, raise CRF or lower the resolution and run it again.

Why the output is always H.264 MP4

VP9 and AV1 produce smaller files than H.264 at the same quality, but they are far slower to encode; AV1 on single-threaded WebAssembly would take an unreasonable time even for a short clip. Playback compatibility differs too. H.264 MP4 plays with hardware decoding on virtually every phone, TV, chat app and older Windows PC, whereas hardware AV1 decoding is limited to relatively recent devices.

The same reasoning explains the x264 preset veryfast. A preset trades compression efficiency for speed: it is faster than the default medium, at the cost of slightly larger files at the same CRF. Audio is re-encoded to AAC at 128 kbps.

Lower the resolution before you raise CRF

Going from 1080p (1920×1080) to 720p (1280×720) leaves 44% of the pixels. On a phone or in a chat preview the difference is hard to see, while the size drops sharply — usually a better result than pushing CRF to 33 until everything smears.

When you pick a resolution, the tool sets the height and derives the width from the aspect ratio, rounding both to even numbers because H.264 cannot encode odd dimensions. Only values smaller than the source make sense. Note that choosing 720p for a vertical 1080×1920 video makes the height 720, which is much smaller than you might expect.

What Keep original and Remove audio actually do

Keep original mode copies the video and audio streams into a new container without decoding them (stream copy). There is no quality loss and it finishes in seconds, but since nothing is recompressed, the size barely changes. Its real use is tidying the container: MP4, WebM, MKV and AVI stay in the same format, while MOV and anything else become MKV. MP4 output has its index (moov) moved to the front (faststart), so once uploaded to the web it starts playing before the whole file has downloaded.

Remove audio also copies the video stream and simply omits the sound. The saving equals the size of the audio track. At 128 kbps that is about 0.96 MB per minute, which makes it a quick fix for screen recordings that never needed sound.

Worked examples

How to use it

1. Choose a video file (above 300 MB you will see a memory warning).
2. Pick a mode; for quality compression, choose the quality step and resolution.
3. On first use the engine (about 30 MB) downloads, then encoding starts.
4. Preview the result, check the size and download.

Frequently asked questions