Merge audio files

Making files agree before they can be joined

A sound file declares one sample rate, one channel count and one codec for its whole length. Files that disagree cannot simply be laid end to end, so the first file sets the terms and anything that does not match is decoded and re-encoded to fit. The practical consequence is worth planning around: joining recordings made the same way — same recorder, same settings, same session — is fast and keeps every file untouched, while joining a 44.1 kHz MP3 to a 48 kHz stereo AAC to a mono voice memo means most of them pay a generation of quality.

Putting the best file first is therefore a real decision rather than a detail. Everything else is brought up or down to match it, so starting with the highest sample rate and channel count in the set avoids converting the good material down to meet the worst of it.

Gapless joins are the hard case, and MP3 is the reason. Every MP3 encoder adds a short run of silence at the beginning and end of a file as a consequence of how it works, and stores the amounts in a header so that good players can skip them. Joining MP3s without accounting for that inserts a small gap at each seam — a few tens of milliseconds, inaudible between spoken segments and immediately obvious across a live album or a continuous mix. If a gapless result matters, join lossless files and encode once at the end.

There are no crossfades between pieces yet, so each file cuts straight to the next. That is fine where the material starts and ends quietly and abrupt where it does not, and a join in the middle of loud continuous sound can produce an audible click for the same reason a cut can.

When you want something else

When the files already match exactly, the concat demuxer in ffmpeg joins them with no re-encoding at all: list them in a text file and run ffmpeg -f concat -safe 0 -i list.txt -c copy out.mp3. It is instant, lossless and has no length limit.

For crossfades, level matching between recordings made at different volumes, or trimming silence from each join, Audacity is free and shows you the waveform while you work — which is the only sane way to assemble anything longer than a few files. For a podcast or an audiobook, that is the right tool rather than this one.