Shrink an MP4, MOV, MKV, WebM or AVI right in your browser — pick a quality level and optionally a smaller resolution, and get a lighter MP4 that still plays everywhere. Nothing is uploaded; your video is compressed on your own device.
In short: drop a video below, pick Balanced or name a size to fit under, and press Compress. It re-encodes to a smaller H.264 MP4 using your device’s own video hardware where the browser exposes it. Free, no watermark, and nothing is uploaded.
Everything happens on your device — nothing is uploaded, so even sensitive footage stays private. Behind the button the tool picks the best encoder your browser offers. First choice is WebCodecs, which drives the same H.264 hardware your phone uses to record video and is not tied to playback speed, so a clip usually compresses in a fraction of its own length. If WebCodecs is missing, it records through that hardware in real time instead — a ten-minute clip then takes about ten minutes. Only when neither is available does it fall back to FFmpeg compiled to WebAssembly, which encodes in software: a one-time ~25 MB download and several minutes for a clip the hardware would finish in seconds. All three produce H.264 in an MP4, and the note under the button says which one ran.
The order matters more than it looks. A failure anywhere in that chain falls through to the next engine rather than failing the job, so a browser that cannot do the fast thing gets a slower conversion instead of an error — and a regression in the fast path costs speed, not your video.
There are two ways to drive it. By quality picks a bitrate for you — Light stays close to the original, Balanced is the sweet spot for sharing, Strong and Extreme trade visible detail for the smallest files. Balanced is the default, with one exception: if your clip is already efficiently encoded — anything re-encoded once already, or exported by an editor — Balanced would barely dent it, so the tool starts you on the first level that actually makes it smaller and tells you it did. To a file size works backwards from a number: name 20 MB for Discord or 25 MB for email and the tool divides that budget across the clip’s length, reserving a slice for audio so the soundtrack never eats the picture.
A size target is only worth something if it is kept, and this is the part most tools get wrong. The first pass is an estimate — how many bits a clip really needs depends on how much it moves, and a static screen recording and a handheld pan differ by several times at the same setting. So the tool measures the file it produced and re-encodes at a corrected bitrate if it came out over, up to two extra passes. If a size genuinely cannot be reached without turning the video to mush, it says so before encoding rather than handing you a smear afterwards.
Two smaller promises come with it. The target is a ceiling, not a quota: a clip that already fits is left alone rather than inflated to fill the number. And the output is capped at the source bitrate, so a compressor can never hand you back something larger than you gave it.
What this tool does and does not do, stated plainly. The “No” rows are the point, not an omission.
| Capability | Supported | Detail |
|---|---|---|
| Input | ||
| MP4, MOV, MKV, WebM, AVI, M4V | Yes | Anything your browser can decode. |
| File size limit | No | There is no server to protect, so nothing is capped by us. |
| Long clips | Yes | Bounded by your device, not by a plan — see the speed section. |
| Several videos at once | No | One clip at a time; the encoder wants the whole device. |
| Controls | ||
| Target file size | Yes | Name a size and the output is measured against it. |
| Quality levels | Yes | Light, Balanced, Strong, Extreme. |
| Resolution downscale | Yes | 1080p, 720p or 480p, or keep the original. |
| Remove the audio track | Yes | A checkbox; useful when a target is tight. |
| Compress more | Yes | One more real step on the same clip, after a weak result. |
| Trimming | No | Use our video trimmer first, then compress. |
| Frame rate control | No | The source frame rate is kept. |
| Crop, rotate or filters | No | This tool only changes size, never the picture. |
| Output | ||
| MP4, H.264 video, AAC audio | Yes | Plays on every device, browser and platform. |
| Never larger than the input | Yes | Capped at the source bitrate, so it cannot inflate. |
| WebM, AV1, HEVC or GIF out | No | H.264 in MP4 only — it is the format everything accepts. |
| Watermark | No | Nothing is added to the picture. |
| How it runs | ||
| Nothing is uploaded | Yes | The clip is encoded on your device and never sent anywhere. |
| Signup or daily limit | No | There is no account and no quota. |
| Offline use | No | The page loads from the network; the software fallback also fetches an engine. |
| API | No | This is a page, not a service. |
Checked 19 September 2026. These move, so treat the date as part of the fact.
| Where you are sending it | Limit | The catch |
|---|---|---|
| Discord (free) | 10–25 MB | Documented at 20 MB, but rolled out in stages — see below. 10 MB always works. |
| Discord Nitro Basic | 50 MB | |
| Discord Nitro | 500 MB | |
| WhatsApp (as video) | 16 MB | Sent as a document instead it goes to 2 GB, but it will not preview. |
| Gmail | 25 MB stated | Really about 18 MB of file: attachments are base64-encoded in transit, which inflates them by roughly a third. |
| Outlook.com | 20 MB | Same base64 inflation — budget about 15 MB of actual file. |
| Slack | 1 GB per file | On every plan. The free plan limits total storage instead, so a big clip is accepted and then eats into it. |
Discord is the one worth reading twice. The free limit is documented as 20 MB, raised from 10 MB in August 2026 — but it has been rolled out to slices of accounts rather than switched on for everybody, so some people still hit 10 MB and some see 25 MB. Most guides print a single number and leave you to discover which one you have when the upload fails. If the clip is for someone else, or you have not checked your own account, compress to 10 MB and it will go through regardless. The presets offer both.
The email numbers have a different catch: attachments are base64-encoded in transit, which inflates them by about a third. Gmail’s stated 25 MB is therefore roughly 18 MB of actual file, and a 20 MB cap is about 15 MB. Aim at the smaller number — that is what the Email preset already does.
Most of it comes from two places. First, phone and screen-recorder videos are stored at a much higher bitrate than they need — the device is encoding in real time with no chance to think about it — so simply re-encoding them carefully shrinks the file a lot with no visible cost. That is why a clip straight off a phone typically loses 40–80% and an already-exported edit loses far less.
Second, dropping the resolution roughly halves the pixel count and the size again. 1080p to 720p is usually invisible on a phone screen, which is where most shared video is watched. If a size target is out of reach, that step buys more room than any quality setting will.
What suffers first is fast motion and fine texture — a handheld pan across foliage, confetti, rain. A talking head, a screen recording or a static product shot survives far harder settings, which is why a single “how much will I lose” number would be misleading and the estimate is shown against your own clip instead.
Nowhere. There is no upload step: the encoding runs in the browser, on your device, and the page contains no code that could send a video anywhere. You can check that in the Network tab — the only requests are for the page itself and, on the software fallback only, the FFmpeg engine from a CDN. Our analytics and ad scripts load as they do on any page and see the page view, never the file.
The practical effects are worth naming, because they are the reason to prefer this over an upload-based compressor: there is no upload wait, which on a long clip is usually most of the time; there is no size cap, because the caps on those sites exist to protect their bandwidth; and nothing of yours is sitting on a server afterwards waiting to be deleted. We are not the only tool that works this way — a handful of browser-based compressors now exist — and the honest trade against a server is that a real CPU can run slower, higher-quality encoder settings than a browser will. For an archival master, a desktop tool like HandBrake still wins.