How In-Browser Video Processing Works

FFmpeg, compiled for the browser

FFmpeg is the open-source engine behind a large share of the world's video software. It can be compiled to WebAssembly, a format browsers run at close to native speed in a sandbox. When you open a tool, your browser downloads that engine (about 10 MB compressed, cached after the first time) and runs it in a background worker so the page stays responsive.

When you drop a file, the browser gives the engine read access to it on your device. The engine writes the result into the tab's memory, and the download button saves it from there. At no point is the file sent to a server.

Some tools do not need FFmpeg at all. The frame grabber and Video to Frames use the browser's own video decoder, and the screen and webcam recorders use the browser's recording features.

How you can check that nothing is uploaded

• Open your browser's developer tools, go to the Network tab, and process a file. You will see the page and engine files load, and no request carrying your video.
• Load a tool page, then turn off your Wi-Fi or mobile data. The free tools keep working, because everything they need is already in your browser.

We do not use analytics cookies or put a watermark on free output, and there are no accounts for the free tools.

The trade-offs: speed and memory

Desktop FFmpeg uses every CPU core and often a hardware encoder in your graphics chip. The browser build used here runs on a single CPU thread and cannot use hardware encoders, so encoding is several times slower than a desktop app. Operations that copy streams, such as fast trim, remove audio and many conversions, are still quick because nothing is re-encoded.

Memory is the other limit. A browser tab can only use so much memory, and phones allow less than computers. Very large files, or tools that hold many frames at once such as Reverse, can run out.

• Prefer fast, stream-copy modes when they fit the job.
• Trim before compressing or converting, so less video is processed.
• Lower the resolution for long re-encodes.
• On a phone, keep re-encoding jobs to shorter clips; the tools warn you when a file is large for your device.

What the Pro AI tools send

Speech recognition, translation, AI voices and vocal separation need large AI models that cannot run well in every browser, so Auto Subtitles, Video Translator, AI Dubbing and AI Vocal Remover use our server. They are designed to send as little as possible.

• Your browser extracts the audio track from the video on your device. For speech tools it is converted to a small mono file first.
• Only that audio track is uploaded. The video itself never leaves your device.
• The audio is stored temporarily for processing and deleted within 24 hours.
• Results such as subtitles or a dubbed voice track come back to your browser, where burning in subtitles or mixing the new voice into your video happens locally.

Each Pro tool says this on the page before anything is uploaded, so you can decide before it happens.