Convert MKV to AAC

A Matroska master built for distribution carries what distribution needs: the original language, one or more dubs, an audio description, sometimes a mix-minus or an interpreter feed. Five audio tracks in one file is entirely ordinary and it is the reason the master is Matroska rather than anything else.

An ADTS file has room for exactly one. That is not a limitation of this tool but of the format: an elementary stream is a run of self-contained frames with a header on each, and there is nowhere in the design to put a second run. So the conversion has to pick, and everything below is about making sure it picks the track your delivery specification named.

When the MKV holds several audio tracks, the first one declared in the file is converted and the rest are discarded. That is file order and nothing else: Matroska’s own "default track" flag is not read, no sorting by language is applied, and the track a player would have started with is not necessarily the one you get.

On material assembled by a ripping or muxing tool the two frequently differ, which is how a commentary or a descriptive audio track ends up in a delivery slot that asked for the original language. Run MediaInfo or `mkvmerge -i` over the file first; both print the tracks in order with their language codes, and it turns a guess into a check.

Two routes, both short. The direct one is to extract the track you want with MKVToolNix — it will pull a single audio track out of a Matroska file without decoding a sample — and convert that single-track file here, at which point there is nothing to choose between.

The indirect one is to convert the MKV to an M4A instead — one of three audio targets on this site that keep every soundtrack, along with Ogg and Opus — and to take the track you need out of that in a tool that lets you select. It costs an extra generation of compression, so it suits a check or a transcript rather than a delivery master.

If the MKV carries tags — a title, a date, a comment written by whatever produced it — they are written into the output as an ID3v2 block ahead of the audio frames. This is a normal thing for a .aac file to contain and essentially every real decoder skips it without comment.

It matters here because this reader’s consumer is often not a real decoder but a script. Anything that assumes byte zero is an ADTS sync word will read the tag as garbage and either reject the file or resynchronise somewhere unhelpful. If the ingest is that kind of parser, clear the tags out of the MKV before converting, or strip the ID3 header afterwards — it is a fixed-length prefix and trivially removable.

Some files stop with a message saying nothing in them could be converted. That is the browser declining to decode the soundtrack, and with the picture discarded there is no other track to fall back on, so the conversion refuses rather than writing an empty file.

The codecs behind it are always the same: AC-3, E-AC-3, DTS and Dolby TrueHD. Those are what disc-sourced masters carry, and no browser ships a decoder for them — DTS and TrueHD are not recognised as codecs at all by the reader. Material from a camera, a capture card, a stream or a download tool is AAC or Opus and is never affected, so the split is entirely predictable once you know where a file came from.

The sample rate and the channel count come across unchanged, so a 48 kHz stereo master produces a 48 kHz stereo stream. That is the answer a delivery specification usually wants, and it is the answer this gives without a setting to get wrong.

There is one fallback worth knowing about. Where a track has a channel layout or a rate the encoder cannot configure, the conversion resamples to standard parameters rather than failing outright — which is the right behaviour for a person and the wrong surprise for a pipeline. Verify the first file of any batch against the spec before running the other ninety-nine.

The audio is decoded and encoded again here, every time, including when the MKV already holds AAC. For a listener that is a shrug. For a master that has already been through a chain of encodes it is another generation stacked on the ones before it, and there is no setting that avoids it.

MKVToolNix and ffmpeg will both copy AAC packets out of a Matroska container into an ADTS stream without decoding a single frame, in seconds, with the result bit-identical to what was inside. Where the file is a deliverable, that is the correct operation and this page is not. Where it is a transcript source, a check, or one file on a machine with nothing installed, this is faster than the download.

The decoding and the AAC encoding both happen inside the browser tab, on your own processor. Nothing carrying the file is sent anywhere, there is no account, and the network tab will confirm it in ten seconds. For unreleased masters under an agreement, that is frequently the difference between a usable tool and one that is out of the question regardless of how good it is.

The free tier accepts an MKV up to 100 MB, measured going in. Delivery masters are usually far past that, which puts most of this reader’s real work on the desktop — an honest limit for a browser tab rather than a nudge towards a subscription. Where the file fits, everything above applies unchanged and nothing is queued behind anyone else.