Convert MP4 to WebM

Most container changes are cosmetic: the same H.264 stream inside a different box. This one cannot be. WebM was defined by Google in 2010 around a specific pair of codecs, and the file this page produces holds VP9 video with Opus audio. The MP4 almost certainly held H.264 with AAC. Nothing is shared, so every frame is decoded and encoded again from scratch.

That is the whole reason this page exists separately from the other container conversions. It is slower than all of them by a wide margin, the settings change the result in ways you can see, and the thing you get in return is a smaller file for the same picture — which only matters if the file is going over a network to somebody else.

VP9 is a generation newer than H.264 and spends fewer bits reaching the same visible quality. For a hero video or a looping background — the two places video actually hurts a page — that difference lands directly on how long a visitor stares at nothing.

It is not guaranteed. Re-encoding an already-compressed MP4 means the VP9 encoder is faithfully reproducing H.264’s artefacts alongside the picture, and on a short clip that has already been squeezed hard the WebM can come out no smaller at all. Look at both file sizes before committing. A conversion that made the file bigger is a conversion to undo.

The `<video>` element takes more than one `<source>` child and the browser uses the first one it can play. Listing the WebM first and the original MP4 second means modern browsers take the smaller file and anything unusual falls back without a single line of JavaScript. Only one of the two is ever downloaded.

Keep the MP4 rather than deleting it once the WebM exists. It costs storage and nothing else, it is the file to hand to someone who asks for a download, and it is the one that will still open in a video editor in ten years. WebM is a delivery format; treating it as your only copy is how a project loses its master.

The three bands map to bitrate targets that the encoder resolves per codec, and balanced is the default. For a muted background loop the small setting is usually right: nobody is watching it closely, it is competing with everything else on the page for bandwidth, and the difference is invisible at the size it plays.

For anything with a face in it, or text, or fine detail somebody is meant to read, balanced is the floor. The high band belongs to files that will be edited or encoded again later, where you are protecting the next generation rather than this one — spending it on a delivery file mostly buys bytes nobody notices.

A smaller frame needs fewer bits to look the same, and the relationship is not gentle: halving the width and height quarters the pixel count. If the video is displayed in a 800-pixel-wide column, serving 4K to it is spending bandwidth on detail the layout throws away before anyone sees it.

The resolution control only goes down. Asking for a size larger than the source has is ignored rather than honoured, because upscaling produces a bigger file containing exactly the detail that was already there. Set it once here rather than letting the browser scale on every playback.

The sound is re-encoded as Opus, which is the audio half of what WebM is for and is genuinely better than AAC at low bitrates — it is what browsers use for voice calls for that reason. On speech at small sizes the difference is audible in Opus’s favour.

The unavoidable cost is that this is a second round of lossy compression on audio the MP4 had already compressed once. For a background loop with no sound, or a demo with a voiceover, this is academic. For a music video where the audio is the point, extract the sound separately from the original at high quality rather than accepting whatever the delivery encode leaves.

WebM is one of the few video formats with an alpha channel, and that reputation causes a specific disappointment here: an MP4 has no alpha to hand over, and any alpha side data present in the stream is discarded during the conversion. A transparent WebM has to be authored from a source that actually holds transparency.

If the goal is a video that sits over a page background without a visible box, the usual answers are a matching solid background colour baked into the video, or a CSS blend mode over an opaque clip. Both are less fragile than alpha video, which several browsers still handle inconsistently.

The conversion asks whether this browser can encode VP9 before it reads a single byte, and if the answer is no it says so and names browsers that can. That check exists so the failure arrives immediately rather than after several minutes of work, phrased as though the file were at fault.

There is deliberately no software fallback. A VP9 encoder compiled to WebAssembly is tens of megabytes to download and roughly an order of magnitude slower than the hardware encoder already present on the machine, and offering that quietly would turn a clear refusal into a conversion that appears to hang.

Video encoding is minutes of processor time, and that is precisely why hosted converters put a watermark on the output, cap the length, or ask for an account: somebody has to pay for those minutes. Here they are yours, so none of that applies and the MP4 is never uploaded.

The practical ceiling is 100 MB on the free tier and, past that, your own memory and patience. On a hosted converter the upload alone would usually take longer than this entire conversion does.