Convert AMR to MP3 — why you probably shouldn't for evidence
Converting an .amr file to MP3 is easy, and it is the first thing most people try. It also makes the file slightly worse, and if the recording is evidence it leaves you holding two files instead of one.
If you have decided you need it, the one-line version with ffmpeg is ffmpeg -i input.amr output.mp3, and to a lossless container instead, ffmpeg -i input.amr -c:a pcm_s16le output.wav. Keep the original either way.
That does not make it always wrong. It makes it a trade, and the trade is usually described backwards.
What the conversion actually does
AMR and MP3 are both lossy. Converting between them decodes the AMR into an approximation of the original audio, then encodes that approximation again under a different scheme that discards its own share.
Nothing is recovered in the process. Detail the AMR encoder removed at capture is gone before the conversion starts, and no MP3 bitrate puts it back — choosing a high one just produces a bigger file holding the same diminished audio.
Why the result sounds about the same anyway
Most of the loss already happened. A narrowband speech codec at capture removes far more than a second-generation re-encode does, so the conversion is rarely dramatic.
That is the honest version of the argument, and it cuts both ways: converting is not a disaster, and it is also not an improvement. You are spending a little audio quality to buy compatibility. Worth doing when you need the compatibility; not worth doing by reflex.
When the trade is worth it
- Something you must use will not accept AMR. A real constraint, and a fair reason. Keep the original alongside the converted copy.
- You need to hand a listenable file to someone who should not have to install a player. Also fair — a working copy for convenience, not a replacement.
And when it is not: as preparation for transcription. There the conversion runs before anything has had a chance to use the detail it is discarding, which is the worst possible moment for it. Submit the original and let one normalization step run from that.
If the recording is evidence
Keep the file you were given, unmodified, and do everything else on copies.
A converted file is a different file: different size, different contents, a different digest. Keeping the original alongside it costs nothing and means you always have the version with the fewest hands on it.
Questions
Does converting AMR to MP3 lose quality?
Yes. Both are lossy formats, so the conversion decodes an approximation and then re-approximates it. Nothing is recovered and a little more is discarded. Choosing a high MP3 bitrate makes the output file larger without putting back what the AMR encoding already removed.
Then why does the converted file sound about the same?
Because most of the loss already happened at capture. That is the point: the conversion is not dramatic, it is just pointless in one direction and slightly harmful in the other. You are paying a small cost for compatibility, not making the audio better.
Is WAV a better conversion target than MP3?
For preservation, converting to 16-bit PCM WAV adds no further loss — but check that is what you are getting. WAV is a container, not a coding scheme, and routinely holds compressed audio such as μ-law, ADPCM or GSM. Either way it does not undo the AMR encoding, and the file gets much larger.
What if I was only given the converted file?
Then you work with what you have, and it is worth knowing that a conversion happened. If the original is obtainable, it is worth obtaining — not because the converted copy is unusable, but because the original is the version with the fewest hands on it.