How to extract audio from a video without uploading it

You have a recorded interview, a client call, a lecture or a screen capture, and you need the audio out of it. The obvious route is a conversion website. The obvious route also means handing a recording of someone's voice to a stranger's server.

For an NDA call or a client interview, that is not a small detail. It is the reason not to.

Your browser already has the decoder

Every modern browser ships with the machinery to decode the audio track of an MP4, MOV, WebM or M4A — it has to, or video would not play. Reading that track out and ignoring the video frames is something the tab can do on its own, with no server in the loop.

The decoded audio is resampled through an offline audio context and written as a RIFF/PCM WAV. Nothing is transcoded remotely, so the quality ceiling is whatever the original recording had.

Which sample rate to choose

• 16,000 Hz mono — what nearly every speech-to-text engine wants, and about a sixth the size of CD-quality stereo. Use this for transcription.
• 44,100 Hz — CD quality. Use it if a human will listen closely or the audio is going into an edit.
• 48,000 Hz — the video standard. Keep it if the audio is going back into a video timeline.
• 22,050 Hz — a middle ground when size matters more than fidelity.

The single most common mistake is exporting a two-hour meeting at 48 kHz stereo and then feeding it to a transcription service that immediately downsamples it to 16 kHz mono anyway. You paid for the size and got nothing for it.

Why there is no MP3 button

MP3 encoding needs a licensed encoder. Shipping one inside a free page is not something we are going to do quietly, so the tool writes WAV, which is uncompressed and readable by everything. If you need MP3 at the end, convert the WAV afterwards with a tool that has the right to encode it.

WAV is roughly 10 MB per minute at CD quality in stereo, and about 1.8 MB per minute at 16 kHz mono. Choose accordingly.

When a file is too long

The whole track is held in memory while it converts. A very long recording can exhaust a tab, and when it does, the page says so rather than hanging and pretending. Dropping to 16 kHz mono cuts the memory need to about a sixth, which is usually enough to get a long recording through.

“There is no audio track”

Two things cause this and the tool distinguishes them. Either the file genuinely has no audio — screen recordings made without a microphone very often do not — or the audio uses a codec your browser cannot decode. Naming both possibilities is more useful than downloading a silent WAV and letting you find out later.