Video to MP3

What actually happens when audio is extracted

A video file is a container holding separate streams — one of pictures, one or more of sound — interleaved so that they play in step. The soundtrack is not woven into the images; it is its own compressed stream, usually AAC in MP4 and MOV files, Opus or Vorbis in WebM, and anything from MP3 to Dolby Digital in MKV and AVI.

Extracting it to MP3 therefore means three things: ignore the video stream completely, decode the audio stream, and encode that sound as MP3 at the bitrate you choose. Skipping the video is what makes this fast. Re-encoding a ten-minute 1080p video would take a browser tab well over an hour; decoding and encoding ten minutes of audio takes seconds to a minute.

The converter is ffmpeg — the same engine behind most desktop video software — compiled to WebAssembly and running inside this page. That is why it reads so many formats, including ones your browser cannot play: AVI, WMV and many MKV files have no preview here, but their audio still converts.

Choosing the bitrate

MP3's size depends only on the bitrate, so you can predict it exactly: 0.96 MB per minute at 128 kbps, 1.44 MB at 192 kbps and 2.4 MB at 320 kbps. A one-hour lecture is about 43 MB at 96 kbps — speech needs no more — while an hour of music at 192 kbps is about 86 MB.

The audio inside most videos was already compressed: phones record AAC at roughly 128–256 kbps, and streaming-quality video carries 96–160 kbps. Encoding the MP3 at a much higher rate than the source cannot add detail that was discarded when the video was made. A sensible rule is to match the source, or go one step above it: 192 kbps covers nearly every phone video.

For anything that is mostly talking — lectures, webinars, interviews, sermons, meetings — choose Mono and 96 kbps. A single channel at 96 kbps sounds cleaner than two channels sharing 96 kbps, and speech has no stereo image worth keeping. For music, keep Stereo and use 192 kbps or more.

Videos with more than one soundtrack

Films and recordings in MKV often carry several audio tracks: languages, a commentary, an audio-description track. When no track is specified, ffmpeg picks the one with the most channels, which is nearly always the main soundtrack — a 5.1 film mix over a stereo commentary.

If you need a different track, open the video in the Video Converter first: it lists every audio track, lets you choose one, and can write an MP4 that contains only that track. Then extract the MP3 from the result here.

Surround audio is downmixed when you choose Stereo or Mono, using ffmpeg's standard mix, so dialogue from the centre channel ends up in both speakers. Leave Channels on Keep original and a 5.1 track stays 5.1 in the MP3 — technically allowed, but many players then only use the front pair, so stereo is the safer choice.

What it does not do

This works on video files you already have. It does not fetch anything from YouTube or other sites, and there is no link box — download sites are a different kind of tool, and most streaming services' terms forbid it. If a lecture or talk you need is available to download from its source, save it first and drop it here.

Protected video cannot be converted. Films bought or rented from Apple, Amazon or similar stores are encrypted, and no legitimate converter can open them.

Very large files have a practical ceiling. The whole video is read into the tab's memory before converting, so a few hundred megabytes is comfortable and multi-gigabyte files may fail on machines with less memory. If that happens, trim the video to the part you need with the Video Trimmer first — trimming in its fast mode copies the streams and takes seconds.

Private by construction

Recordings of meetings, lectures and family events are exactly the files people should not have to upload to a stranger's server to extract a soundtrack. Here there is no upload step: the video is read from your disk, processed by ffmpeg in this tab, and the MP3 is offered as a download straight from memory.

The one thing the page downloads is the engine itself — about 32 MB, the first time you convert, kept for the rest of the session. After that you can disconnect from the network and convert as many files as you like.