Skip to content
CreatorPrivacyKit

Compress video on your device, not in the cloud

Compress video on your device, not in the cloud?

Use the compressor when you need a smaller copy for a size cap. Compare the exported bytes with the input. A size target is the upper size you set. Compare the download with the input. The page shows the byte change.

Gmail Help: Send attachments with your Gmail message; WhatsApp Help: media and documents; WhatsApp Help: longer videos; Telegram FAQ

Last updated

Use the compressor when you need a smaller copy for a size cap. Compare the exported bytes with the input. A size target is the upper size you set.

Reviewed 9 October 2026

Check the file you actually got

Compare the download with the input. The page shows the byte change.

Set resolution, quality, or a megabyte target, then compare the same frame and the exported bytes.

Auto output is not always MP4: VP9 or AV1 writes WebM. Encoding runs in this tab with WebCodecs.

2026-10-09 single-sample production test

Three rows from one 10-second clip, recorded on 2026-10-09.

Anonymous clip: 10 seconds, 720×1280, 24 fps, 240 frames, H.264, no audio track, 9,491,383 bytes. All three runs kept Original size and FPS, Auto codec, quality 50.

Preset Size target Output Output bytes Change SSIM Same-frame note
Balanced Off H.264 1439264 −84.84% 0.898303 Same-frame comparisons showed damage to hair, accessories, and fabric.
Balanced On, 5 MB upper target H.264 4071380 −57.10% 0.955849 Best detail of these three on this sample, still lossy. 5 MB is an upper target, not a precise output promise.
Smallest Off VP9 WebM 949127 −90.00% 0.881243 Detail loss.

SSIM in this table is the score recorded for this clip.

All three outputs decoded 240 frames and the full 10 seconds. Matching frames were compared. No full real-time viewing or listening acceptance was performed. The original had no audio.

Rounded timing from that run

The UI showed about 1 / 1 / 2 seconds. Those readings are rounded. This run did not record an exact Chrome version.

Open the labs compression report

The labs compression report lists the 2026-10-26 color-bars export, 2976898 bytes to 149460 bytes.

The math: how file size actually works

Average video bitrate, audio bitrate, and duration largely determine the bytes. Mux and container overhead, plus encoder rate-control variation, still move the actual size away from a simple product.

Lowering resolution alone does not reliably reduce size when the target bitrate stays similar. A lower resolution can look sharper at a lower bitrate, rather than blocky, but the bytes still follow the bitrate budget, overhead, and rate control.

Audio bitrate cannot be ignored on tight limits: uncompressed or high-bitrate audio takes significant space. A 192 kbps stereo track consumes ~1.44 MB per minute. For 20 MB or 25 MB caps, dialing audio to 96–128 kbps AAC reclaims essential budget for video clarity.

Platform upload limits (2026 verified)

Upload caps are enforced at the transport boundary before playback:

Containers and Codecs

Privacy and metadata

Raw camera footage can embed camera model, lens serials, creation timestamps, QuickTime udta user records, and GPS coordinates.

Re-encoding writes a new container and drops supported container tags. Pixel-domain watermarks stay. Inspect the downloaded file if you need to confirm what remains.

To check that the media file is not posted to this site: open browser DevTools (F12), switch to the Network tab, filter by Fetch/XHR, and compress a file. Other page requests can still appear.

Workflows and platform presets

Related: Private video compression · Methodology.