Convert any video to MP3

Yes. Wave loads a WebAssembly build of FFmpeg into the page and runs the whole extraction locally. You can open your browser's network tab while converting and confirm - no file bytes leave the tab.

Every video source Wave supports has an MP3 extract page: MP4, MKV, MOV, WEBM, AVI, WMV, FLV, TS, MTS, M2TS, 3GP, 3G2, MPEG, VOB, OGV, F4V, RM/RMVB, DivX, M4V, MXF, DV, AV1, and ASF. Sources with no audio track (raw H.264/HEVC/VC-1 elementary streams, GIF) are rejected during inspect with a clear message.

One format on purpose. MP3 plays on every consumer device that plays sound - phones, hardware players, car stereos, smart speakers, voice assistants, DAWs. FLAC and WAV are lossless but ~10x larger and don't universally play (many car stereos and older phones lack decoders). Opus is more efficient than MP3 but has partial player support outside browsers. MP3 is the pragmatic universal target.

192 kbps VBR through LAME (libmp3lame -q:a 2). That falls inside the "transparent to trained listeners on studio monitors" range for essentially all music and voice content. Going higher (256, 320) yields no perceptible improvement on consumer playback; going lower (128 and below) starts showing artefacts on cymbals, sibilants, and applause.

Much faster. Wave discards the video track entirely - it never decodes a single frame - so the conversion runs at 3-10x the speed of an equivalent video re-encode of the same file. An hour-long source typically produces an 85 MB MP3 in 20-40 seconds on modern hardware.

A copy path (bit-exact stream copy) is possible when the source audio is already MP3 - which happens on many AVI, FLV, MKV, and DivX sources. In that case Wave moves the audio bytes into an MP3 file without decoding a sample, and the output is byte-for-byte identical to the source audio. Re-encode is what happens for everything else: decode to PCM, feed to LAME, get MP3 out.

No - and it rejects during inspect with a friendly message before conversion runs. Sources like raw H.264/HEVC elementary streams, GIF, and some silent OGV files have no audio and cannot produce an MP3.

Wave downmixes multi-channel audio to stereo before the MP3 encode. AC-3 and DTS sources carry explicit downmix metadata (Dolby Lt/Rt and Lo/Ro coefficients) that Wave respects, so dialogue stays centred and effects preserve approximate stereo position. Object-based audio (Dolby Atmos, DTS-X) folds to stereo without the height layer.

Basic tags (title, artist if present) transfer to the MP3's ID3v2 frame. Album art doesn't transfer through the current pipeline - use a desktop tag editor if you need artwork.

Yes. After you drop the file, open Options and type a Start and End time under Trim, like 1:30 and 3:00; Wave encodes only that part to MP3. Times take tenths of a second too (0:42.5), and leaving End empty runs to the end of the file.

The default-flagged track. Broadcasters and Blu-ray authors set this to the primary language's main mix; media containers with alt-language dubs or director commentary tag those tracks as non-default. Multi-track extraction is on the roadmap; for now, `ffmpeg -map 0:a:N` picks a specific stream on the command line.

Up to 8 GB per source file. Wave streams the source through the WebAssembly engine via WORKERFS in chunks, so the ceiling is your browser's decoder headroom, not RAM. Above 2 GB the widget shows a heads-up so you know the extract will take a moment.

Free. No sign-up, no trial, no per-file limit, no watermark on the output. The site is supported by nothing external - it runs on Cloudflare Pages and the work happens on your device.