Convert MKV to OGG

Lecture and conference recording systems write Matroska more often than anything else. Opencast produces it, OBS records to it by default because an MKV survives a crash where an interrupted MP4 does not, and camera-plus-slides capture rigs use it because they need to keep two video streams and several microphones in one file without arguing with a container about it.

That is excellent for recording and unhelpful for publishing. The systems these recordings end up in — institutional repositories, teaching platforms, wikis, federated video hosts — have accepted-format lists written around what is free to implement, and Matroska sits awkwardly across that line while Ogg sits squarely on the right side of it.

Ogg is a container, not a codec, and that distinction is the source of most of the trouble on this pair. It has carried Vorbis since 2000, FLAC since not long after, and Opus since 2012. A file with a .ogg extension therefore tells you almost nothing about what a decoder will need.

What this conversion writes is Opus. It is the right default by a wide margin — better sound per byte than Vorbis at every bitrate that matters, and specifically designed for speech, which is what a lecture is. It is also the answer that will surprise somebody whose requirement was written a decade ago, because for most of Ogg’s life "Ogg" meant Vorbis and some software still assumes it.

The accepted-formats list on a repository is frequently older than the software serving it, in both directions: a page that says "Ogg Vorbis" is often backed by a player that handles Opus perfectly, and a page that says "Ogg" is occasionally backed by one that does not. Neither can be resolved by reading.

So convert one short recording, upload it, and play it back through the destination’s own player rather than your own. Ten minutes spent on that at the start of a term is the difference between a working archive and eighty files that all need doing again. If Opus turns out not to be accepted, the conversion you need is to Vorbis, and that has to come from a desktop tool — nothing here writes it.

MP3, AAC, WAV and FLAC each hold exactly one soundtrack, so converting a multi-track MKV to any of them throws the rest away. The Ogg container has no such limit, and this conversion writes every audio track the browser can decode into the file rather than choosing one.

Treat that as storage rather than as a feature you can rely on. Multiplexed Ogg is legal and not much exercised, so whether a particular player offers you the second stream is a question about that player and worth testing before it matters. A lecture capture with a lectern microphone and a room microphone will contain both either way; if switching between them is the point, the M4A target is the one designed for it.

A capture from OBS or a merged download frequently carries Opus already, which makes it tempting to assume the stream is lifted across. It is not: every track is decoded back to samples and compressed again, so Opus to Opus costs a second generation.

On speech recorded in a hall this is genuinely inaudible, and for the volume of material a teaching archive produces the convenience is worth more than the difference. Where it is not — a recorded concert, a music seminar with performance in it — ffmpeg or a desktop remuxer will copy the existing Opus stream into an Ogg container in seconds with nothing changed at all.

The size that matters is not the size of the MKV. The picture is discarded rather than converted, and in a lecture recording the picture is nearly all of the file — a fixed camera on a speaker for ninety minutes compresses badly and stores enormously. What remains is an audio file whose size follows the quality band and the duration and nothing else.

Three bands are offered, defaulting to balanced. For a single voice in a room the small band is defensible and roughly halves the result; for a panel, a discussion or anything where several people talk across each other, balanced is worth its size. Opus is efficient enough that the difference between the two on a term of recordings is measured in tens of megabytes rather than gigabytes.

If a file stops with a message saying nothing in it could be converted, the browser could not decode the soundtrack and, with the picture already discarded, there was nothing left to write. The codecs involved are AC-3, E-AC-3, DTS and Dolby TrueHD.

A lecture capture will essentially never hit this, because capture software writes AAC or Opus. It is the disc-sourced Matroska file that fails — which is worth knowing if the archive you are publishing includes anything that came from a DVD of an older event. Those need their audio converted to AAC once by a desktop tool before anything here will read them.

The extension is a convention rather than a property. Xiph suggests .oga for audio in an Ogg container and .opus for Opus specifically, and most of the world carried on using .ogg for everything. If the upload form names .ogg, use .ogg; renaming a file changes nothing inside it and nothing here depends on the name.

The conversion itself happens inside the browser tab, on your own processor, with no account and nothing sent to a server. That matters for teaching recordings more than it does for most material: a lecture capture contains students’ voices, questions and sometimes names, and uploading it to a converter to get an audio file out is a data transfer somebody would have had to think about. Here there is nothing to think about, and the free tier takes files up to 100 MB.