Video to Audio

At a glance

What actually changes

A video file is a container, not a single stream. An .mp4 holds a video track, one or more audio tracks, and often subtitles and chapter data, all interleaved. Extracting audio means demultiplexing that container — separating the streams — and keeping only the audio.

What the audio actually is depends on where the video came from. In an MP4 or MOV it is almost always AAC. In a WebM it is usually Opus or Vorbis. In an MKV it could be any of those, or AC-3, or something unusual. This tool decodes whichever it finds and re-encodes it to the format you pick.

That re-encode is worth understanding: because the audio inside the video was already compressed, converting it to MP3 is a second lossy pass, exactly as it would be converting a standalone AAC file. Choosing WAV avoids a second lossy encode, at the cost of a much larger file.

When you'd want this

• Pulling audio out of a recorded meeting or lecture so it can be transcribed — video is unnecessary weight for that, and most transcription services charge by duration regardless. The transcription prep tool handles the format side.
• Lifting a soundtrack, interview or podcast recording out of footage you shot yourself.
• Salvaging audio from a screen recording — QuickTime, OBS and Loom all produce video files where the audio is the only part you wanted.
• Getting a track off a video you made to reuse in an edit, in CapCut or DaVinci Resolve.
• Anything confidential. Because nothing is uploaded, an internal recording or a client's footage never leaves your machine. Most online extractors upload the entire video to a server first, which for a private recording is the whole problem.

Settings that matter here

Pick MP3 at 192 kbps or higher for speech and general use — that is well above the point where the second encode is noticeable on dialogue. For music lifted from a video, use 320.

Pick WAV if the audio is going into an editor for further work. It skips the second lossy encode entirely, so nothing is degraded before you start, and every DAW and NLE imports it without complaint. The file will be large — roughly 10 MB per minute — but that rarely matters for a working file.

Expect the first conversion to take a moment. Video containers are decoded by a WebAssembly build of FFmpeg that is fetched on first use; after that it is cached and subsequent conversions start immediately. Larger videos also simply take longer, because the whole file has to be read and demultiplexed before any audio comes out.

What you lose

The video, obviously — but also anything else in the container. Subtitle tracks, chapter markers and timecode are discarded, and if the file has multiple audio tracks (a second language, a separate commentary, or discrete channels from a multi-mic shoot), you will get the first one rather than a choice.

You lose a little quality on any lossy output, for the double-encode reason above. Choosing WAV avoids this.

Very large files are the practical limit. Browsers cap how much memory a page can allocate, and the whole file has to be held while it is processed, so somewhere past roughly 500 MB — a long recording at high bitrate — conversion will fail rather than run slowly. Splitting the source video first is the workaround.