Convert MKV to WAV

OBS Studio writes Matroska by default, and it is a deliberate choice: an MKV survives a crash because its structure is written continuously, where an interrupted MP4 is often unreadable. Anyone recording a podcast, an interview or a stream with OBS therefore ends up with an MKV whether or not they ever intended to make a video.

The sound is usually the deliverable and the picture is a byproduct. That is the shape of this conversion — not extracting audio from a film, but retrieving the recording from the safety container it was written into so that an editor can open it.

This is the part that costs people an afternoon. OBS routes a microphone, desktop audio, a browser source and a voice chat to separate tracks in the same recording precisely so that they can be balanced independently later. A WAV file has room for exactly one audio stream, and when the MKV holds several the first one declared is used and the rest are discarded silently.

The workaround is short and it is worth taking. Convert the MKV to a MOV or an MP4 instead: both carry every audio track the browser can decode, in the original order, and every DAW and NLE will import that file and present the tracks separately. Do the splitting there, export the one you want as a WAV from inside the session, and nothing is lost on the way.

PCM stores every sample it is handed, so this conversion adds nothing of its own. That is a real property and it is the reason to use WAV before destructive editing: noise reduction, compression and equalisation all work on the samples, and each pass through a lossy codec between edits compounds what the previous one approximated.

It is not the same as recovering the original recording. The audio inside the MKV was compressed by AAC or Opus when it was written, and the WAV preserves exactly what those codecs decode to — including the detail they removed at the time. Nobody can put that back. What the WAV guarantees is that nothing further is lost between here and the edit.

The output is 16-bit PCM at the source’s own sample rate, which for anything recorded alongside video is almost always 48 kHz. Sixteen bits rather than twenty-four is worth knowing if you are used to field recordings: it is CD depth, roughly 96 dB of dynamic range, and it is what the compressed source decodes to in the first place.

That is why there is no depth control and no quality band on this page. Widening a 16-bit source to 24 adds zeros rather than resolution, and there is no fidelity-against-size trade to offer because PCM does not compress. Set your session to 24-bit if the project calls for it, and import the WAV into it — the conversion and the session depth are separate questions.

Uncompressed stereo at 48 kHz and 16 bits is 192 kilobytes every second: about 11.5 MB a minute, or roughly 690 MB for an hour of talking. There is no variation and no surprise — the size is a multiplication, which is one of the few pleasant things about the format.

For working files that is fine and for storage it is not. FLAC holds exactly the same samples in roughly half the space, verifies them with a checksum, and is read by Audacity, Reaper, Audition and every archive tool worth using. The sensible split is WAV while the file is being worked on and FLAC once it is finished, and both are available from the same MKV here.

Some files stop with a message saying nothing in them could be converted. That is the browser declining to decode the soundtrack: with the picture already discarded, there was no track left to write, so the conversion refuses rather than producing an empty file.

It is always the same small set of codecs — AC-3, E-AC-3, DTS and Dolby TrueHD, which is disc audio and which no browser decodes. It never affects a recording from OBS, a capture card, a phone or a download tool, because those write AAC or Opus. If you are working with disc material, convert its audio to AAC once with a desktop tool and everything afterwards behaves normally.

Audacity, Audition and most DAWs read audio files rather than video containers, and the ones that read video are looking for MP4 or MOV. Matroska sits outside that set for the same reason it sits outside Final Cut’s import list: it is permissive enough that supporting it means supporting everything it can legally hold.

Some tools will open an MKV through an ffmpeg plugin and it is worth trying before converting, because a direct import saves a step and a generation. Where it works, use it. Where it does not, this conversion is the same operation done in the browser rather than by a library the editor may or may not have shipped.

Everything happens in the browser tab on your own processor. Nothing carrying the recording is sent anywhere, which for an unpublished interview or a client session is often not a preference but a requirement — the ordinary alternative is uploading the whole thing to a site you found a minute ago.

The free tier takes files up to 100 MB, measured on the MKV going in. That is an awkward ceiling here, because a long recording is mostly picture: an hour of talking over a static shot can be a gigabyte of video wrapping forty megabytes of sound. Where the file is too large, trimming it first or using a local tool that copies the audio stream out without decoding is the shorter path.