Built for Speech, Not Music
AMR uses ACELP (Algebraic Code-Excited Linear Prediction), which models the human vocal tract rather than encoding the waveform generally. Speech survives; music, ambience, and anything outside the voice band is mangled. The "adaptive multi-rate" name refers to its ability to switch bitrates on the fly as network conditions change during a call.
Technical Characteristics
Classic AMR (AMR-NB, narrowband) samples at 8 kHz and offers eight bitrates from 4.75 to 12.2 kbps- A three-minute recording can be under 300 KB. AMR-WB (wideband, the .awb extension) samples at 16 kHz at up to 23.85 kbps and is the codec behind "HD Voice" calls. Both are mono by design. The telephone-grade frequency range (300-3,400 Hz for narrowband) is why AMR recordings have that unmistakable phone-call sound. Frames are 20 ms long, and the codec includes error concealment and comfort-noise generation for dropped frames- Features built for unreliable radio links rather than file playback.
When You Meet AMR Files
Voice recorder apps on older Android and feature phones, MMS voice clips, and call-recording tools commonly produce .amr files. There is rarely a reason to create new AMR files today- Modern phones record voice memos as M4A/AAC, and Opus beats AMR at every bitrate. The task is almost always getting audio out of AMR, not into it.
Why Converting Is Worth It
Converting cannot add quality that was never captured- An 8 kHz narrowband file will still sound like a phone call as an MP3- But it future-proofs the recording. MP3 plays in every app and browser, is accepted by transcription services and editing tools that reject .amr uploads, and can be merged into compilations alongside other audio. If a recording has legal or sentimental value, keep the original AMR too; it remains the primary source.
Compatibility and Converting on mp3.now
Desktop support for AMR is poor- Windows Media Player, many Mac apps, and most web players will not open it, which strands old recordings. The AMR to MP3 converter fixes that: upload the .amr file and download an MP3 that plays anywhere. Since the source is low-bitrate mono speech, a modest MP3 setting like 128 kbps is more than enough- Higher bitrates only enlarge the file without improving a thing. Afterwards you can trim or normalize the recording for clarity.