Convert AAC to MP3

This is the fact the rest of the page hangs on. AAC as standardised in ISO/IEC 13818-7 is a codec, and the usual way to store it is inside a container — an MP4, an M4A, a Matroska file — that records how long the audio is, where each frame starts and what the track is called. A file with a bare `.aac` extension normally skips all of that: it is ADTS, an unbroken run of audio frames each carrying a seven-byte header, designed for broadcast where a receiver may tune in halfway through and must be able to start decoding from wherever it lands.

That design is excellent for its purpose and awful as a file on a disk. There is no field anywhere that says how long the audio is, so a player either estimates from the file size — which is wrong the moment the bitrate varies — or shows nothing. There is no index, so seeking means guessing a byte offset and hunting for the next frame header, which is why scrubbing jumps and why the progress bar drifts.

Almost nobody creates one deliberately. They arrive from streaming: HLS delivers audio as a sequence of `.aac` segments and joining them end to end produces exactly this, which is what a download helper or a stream recorder leaves behind. IP cameras, dashcams, DVR boxes and broadcast equipment write ADTS for the same reason radio does — it can be cut anywhere. Some Android recording apps and a few command-line pipelines default to it.

The consequence is that the person reading this rarely chose AAC and rarely cares about it. They have a file that will not behave, and the question is what to turn it into. That framing is worth keeping, because it changes which of the two obvious answers is right.

The MP3 written here opens with a Xing frame holding the total frame count, the file size and a hundred-entry table of contents. That is what lets a player state the duration immediately instead of estimating it, and it is what makes dragging the progress bar land where you dropped it even when the bitrate varies across the file.

On top of that it carries ID3v2 tags, which every player, phone and car system reads for the title, the artist and the cover image. None of that is a property of the MP3 codec — it is the file layout around it. That is the actual upgrade you are buying, and it is why converting fixes problems that "AAC is less compatible than MP3" does not explain.

Every problem in the section above is a container problem, and MP3 is not the only container that solves it. Converting the same file to M4A wraps the audio in MP4, which gives it a duration, a proper sample table for seeking, and a metadata block that holds tags and artwork — and does it without a second lossy encode, because the audio only has to be moved rather than re-imagined.

So pick MP3 for one reason and one reason only: something downstream requires MP3 specifically. A car head unit indexing a USB stick, a player made before 2010, a submission form that lists a single format, a broadcast automation system with an MP3 profile. If your goal is just "make this file work properly on my computer", M4A is the better answer and this page has done its job by saying so.

The registry flags generation loss as a known problem of AAC, and it is the right flag: this conversion decodes audio that has already been compressed and compresses it again with a codec that makes different decisions about what to throw away. The second encoder spends bits describing the first one's artefacts, which is why two lossy passes cost more than one at the same bitrate.

How much you should care depends on what the file is. A recorded stream of speech, a lecture, a broadcast capture or camera audio at Balanced — 160 kbps — is indistinguishable in any realistic listening situation. Music captured at a low stream bitrate is where a second pass shows, most audibly on cymbals and crowd noise. The defence is to convert once from the original capture, never from an MP3 you made earlier.

With a stream capture you usually do not know what bitrate the AAC was encoded at, and the file gives you no easy way to find out. The safe rule is not to try to match it. Small file writes 96 kbps, Balanced writes 160 and High quality writes 320, and those are the only three numbers involved.

For speech, Small file is genuinely sufficient and turns a three-hour capture into roughly 130 MB. For anything musical, Balanced. Choose High quality when the MP3 is a working copy that will be trimmed, normalised or exported again — 320 kbps recovers nothing, but it stops this step from being the weakest link in a chain of edits.

Broadcast and camera AAC is frequently not what a music file would be. It can be mono, it can be at 32,000 or 48,000 Hz, and in the case of a television capture it can carry a multichannel layout — the format allows up to 48 channels, against MP3's two.

Mono and unusual sample rates convert without incident and stay as they are. A layout no MP3 encoder will accept triggers a fallback to two channels at 48,000 Hz, so a surround capture is folded down to stereo and resampled rather than refused. That is the right behaviour for the person who wanted a file that plays, and the wrong one if the discrete channels were the point.

Two failures look alike and are not. If the file is genuinely ADTS but truncated mid-frame — the common result of a stream recorder being stopped abruptly — the conversion reads what is there and stops at the damage, which usually gives you a usable MP3 slightly shorter than the original.

If the extension is wrong, nothing works. Files named `.aac` that are actually MP4 fragments, or M4A files renamed by hand, will not parse as an ADTS stream, and the converter says the file could not be read rather than guessing. Try the M4A route in that case: an MP4 demuxer will open it, and you will have found out what you actually had.

Recordings of this kind arrive in bulk, so the batch path matters. Drop up to a hundred files at a time; each converts independently and the results come back as one ZIP, with a 100 MB ceiling per file. Nothing waits in a queue, because the work is happening on your own processor rather than on a server we would have to pay for.

What this does not do is join them. Fifty HLS segments converted here become fifty MP3s, not one continuous recording. If the segments are pieces of a single stream, concatenate them into one `.aac` first with a tool that can do it, then convert that — the frame layout of ADTS makes plain concatenation valid, which is one of the few genuine conveniences of the format.