Learn how to convert AMR voice recordings to MP3. Step-by-step guide for recovering old voicemail, phone recordings, and WhatsApp voice notes stored in the AMR format.
If you have stumbled across a folder of .amr files on an old phone backup, a USB drive from 2012, or buried in a WhatsApp data export, you are not alone. AMR (Adaptive Multi-Rate) files are a relic of the mobile phone era before smartphones standardized on AAC and Opus — and they are surprisingly difficult to play on modern devices.
AMR was the dominant format for mobile voice recordings, voicemail systems, and MMS audio messages from roughly 2003 to 2015. Nokia, Samsung, Sony Ericsson, and most early Android phones used AMR as their default voice recorder format. Older versions of WhatsApp stored voice notes as AMR files. Many corporate voicemail systems still deliver messages in AMR format.
The problem is that AMR support on modern devices is inconsistent at best. Windows does not play AMR natively. macOS can sometimes handle it, sometimes not. Most music players and audio editors ignore AMR files entirely. If you want to listen to, share, or archive these recordings, you need to convert them to MP3 or another widely supported format.
This guide covers everything you need to know: what AMR actually is, how to convert it to MP3 with the best possible quality, how to handle batch conversion of large collections, and how to rescue audio quality from a format that was never designed for high fidelity.
| Property | AMR-NB (Narrowband) | AMR-WB (Wideband) |
|---|---|---|
| Full Name | Adaptive Multi-Rate Narrowband | Adaptive Multi-Rate Wideband |
| File Extension | .amr | .awb or .amr |
| Sample Rate | 8,000 Hz | 16,000 Hz |
| Bitrate Range | 4.75 - 12.2 kbps | 6.6 - 23.85 kbps |
| Channels | Mono | Mono |
| Frequency Response | 300 - 3,400 Hz | 50 - 7,000 Hz |
| Compression | Lossy (ACELP) | Lossy (ACELP) |
| Primary Use | Voice calls, voicemail | HD Voice calls |
| Quality | Telephone quality | FM radio quality |
The key thing to understand about AMR is that it was designed exclusively for voice at extremely low bitrates. At its most common setting (12.2 kbps for AMR-NB), the entire audio signal is compressed into about 1.5 kilobytes per second. For comparison, a low-quality MP3 at 64 kbps uses roughly 8 kilobytes per second — more than 5 times the data rate. AMR achieves this extreme compression by being optimized for the narrow frequency range of human speech and discarding everything else.
This means AMR files have an inherent quality ceiling. No amount of conversion can make an 8 kHz, 12.2 kbps AMR recording sound like a studio recording. But converting to MP3 makes the files playable everywhere and allows you to apply enhancements that improve clarity.
.amr files (newer versions use Opus)# Simple AMR to MP3 conversion
ffmpeg -i recording.amr -c:a libmp3lame -b:a 128k -ar 44100 recording.mp3
# With quality optimization for voice
ffmpeg -i recording.amr -c:a libmp3lame -b:a 96k -ar 22050 -ac 1 recording.mp3
Since AMR source audio has a maximum frequency content of 3,400 Hz (narrowband) or 7,000 Hz (wideband), using a high sample rate like 44.1 kHz is technically unnecessary. The extra samples above the source's frequency ceiling contain silence. However, 44.1 kHz ensures maximum compatibility with all players and devices, and the file size difference for short recordings is negligible.
# Enhanced conversion with noise reduction and normalization
ffmpeg -i recording.amr \
-af "highpass=f=80,lowpass=f=4000,anlmdn=s=7:p=0.002:r=0.002,loudnorm=I=-16:TP=-1:LRA=11" \
-c:a libmp3lame -b:a 128k -ar 44100 recording.mp3
This pipeline:
Pro Tip: AMR files from voicemail systems often have very low volume. Always apply loudness normalization during conversion. A recording that was barely audible as an AMR file can become perfectly clear as a normalized MP3. Our full audio normalization guide covers all the normalization options.
If you do not want to install FFmpeg, our audio converter handles AMR to MP3 conversion directly in your browser:
For format-specific conversion pages, use the MP3 converter hub.
If you have a folder full of old voice recordings, batch conversion saves significant time.
#!/bin/bash
# Convert all AMR files in a directory to MP3
for amr in /path/to/recordings/*.amr; do
filename=$(basename "$amr" .amr)
ffmpeg -i "$amr" \
-af "highpass=f=80,loudnorm=I=-16:TP=-1:LRA=11" \
-c:a libmp3lame -b:a 128k -ar 44100 \
"/path/to/output/${filename}.mp3" -y
done
echo "Conversion complete!"
# Convert all AMR files in a directory to MP3
Get-ChildItem -Path "C:\recordings" -Filter "*.amr" | ForEach-Object {
$output = "C:\output\$($_.BaseName).mp3"
ffmpeg -i $_.FullName -af "highpass=f=80,loudnorm=I=-16:TP=-1:LRA=11" -c:a libmp3lame -b:a 128k -ar 44100 $output -y
}
AMR files from old phones often have meaningful creation dates (when the recording was made). By default, the converted MP3 will have today's date. To preserve the original timestamp:
#!/bin/bash
# Convert AMR to MP3 while preserving original file timestamps
for amr in /path/to/recordings/*.amr; do
filename=$(basename "$amr" .amr)
output="/path/to/output/${filename}.mp3"
ffmpeg -i "$amr" -c:a libmp3lame -b:a 128k -ar 44100 "$output" -y
# Copy the original file's modification timestamp
touch -r "$amr" "$output"
done
For large batch operations, our batch processing guide covers additional optimization techniques.
AMR recordings are inherently low quality, but several post-processing techniques can improve clarity and listenability.
Old phone recordings often have significant background noise — hiss, hum, wind noise, or ambient sounds:
# Aggressive noise reduction for very noisy recordings
ffmpeg -i recording.amr \
-af "anlmdn=s=10:p=0.003:r=0.003,highpass=f=100,lowpass=f=3500" \
-c:a libmp3lame -b:a 128k recording.mp3
The anlmdn filter uses non-local means denoising, which is effective at removing steady-state noise (hiss, hum) without distorting the voice. The s parameter controls strength — higher values remove more noise but can introduce artifacts.
Many AMR recordings have extreme volume variations — the speaker moves closer and further from the phone's microphone:
# Compress dynamics and normalize
ffmpeg -i recording.amr \
-af "acompressor=threshold=-20dB:ratio=4:attack=10:release=200,loudnorm=I=-16:TP=-1:LRA=7" \
-c:a libmp3lame -b:a 128k recording.mp3
Boosting the frequencies most important for speech intelligibility can improve clarity:
# Speech clarity EQ: boost presence (2-4 kHz) and cut mud (200-400 Hz)
ffmpeg -i recording.amr \
-af "equalizer=f=300:t=q:w=1:g=-3,equalizer=f=3000:t=q:w=1:g=4,loudnorm=I=-16:TP=-1" \
-c:a libmp3lame -b:a 128k recording.mp3
Combining all enhancements for maximum clarity:
ffmpeg -i recording.amr \
-af "highpass=f=80,\
lowpass=f=4000,\
anlmdn=s=7:p=0.002:r=0.002,\
equalizer=f=300:t=q:w=1:g=-3,\
equalizer=f=2500:t=q:w=1.5:g=4,\
acompressor=threshold=-20dB:ratio=3:attack=10:release=200,\
loudnorm=I=-16:TP=-1:LRA=7" \
-c:a libmp3lame -b:a 128k -ar 44100 enhanced.mp3
Pro Tip: When enhancing old recordings, process a short sample first and compare it to the original before running the full batch. Over-processing (too much noise reduction, too much EQ boost) can make voice sound robotic or tinny. Find settings that improve clarity without artifacts, then apply to all files.
WhatsApp voice notes are stored in:
/sdcard/WhatsApp/Media/WhatsApp Voice Notes/The voice notes are organized in folders by date. File names follow the pattern PTT-YYYYMMDD-WA0001.amr (older versions) or .opus (newer versions).
Old phone backup software (Nokia PC Suite, Samsung Kies) stored recordings in proprietary backup formats. You may need:
.nbu or .nbf backup files using NbuExplorer (free, open source).amr files, often under Sounds/ or Voice Recorder/Many voicemail systems allow you to forward messages as email attachments. The attachments are typically .amr or .wav files. If your system supports it, forward all important voicemails to your email before decommissioning the service.
| AMR Source | Recommended MP3 Bitrate | Sample Rate | Channels | Reasoning |
|---|---|---|---|---|
| AMR-NB (8 kHz) | 64-96 kbps | 22.05 kHz | Mono | Source has minimal frequency content |
| AMR-NB (archival) | 128 kbps | 44.1 kHz | Mono | Maximum compatibility, slight overhead |
| AMR-WB (16 kHz) | 128 kbps | 44.1 kHz | Mono | Wideband has more frequency content |
| For sharing | 128 kbps | 44.1 kHz | Mono | Universal compatibility |
| Minimum viable | 48 kbps | 22.05 kHz | Mono | Smallest file, acceptable for voice |
For AMR source material, encoding to MP3 above 128 kbps provides zero benefit. The source audio simply does not contain enough frequency content to justify higher bitrates. The encoding is not adding quality — it is just preserving what the AMR file already contained in a more compatible container. For understanding how bitrate affects audio quality generally, see our audio bitrate quality guide.
While MP3 is the most universally compatible choice, other formats may be better for specific use cases:
For a broader comparison of audio formats, see our AAC vs MP3 comparison and OGG vs MP3 comparison.
AMR is a mobile telephony format with limited support on desktop operating systems. Windows Media Player and many music apps do not include an AMR codec. VLC media player can play AMR files (it includes its own codecs), but for long-term compatibility, converting to MP3 is the better solution.
No. The conversion changes the container format but cannot add audio information that was not in the original recording. An AMR file recorded at 12.2 kbps with 8 kHz sample rate has a fundamental quality ceiling. However, post-processing (noise reduction, EQ, normalization) during conversion can improve the perceived clarity of the voice. The raw audio data is the same, but it can be made more listenable.
Extremely small. A 1-minute AMR-NB recording at 12.2 kbps is approximately 90 KB. The same duration as a 128 kbps MP3 is approximately 960 KB — roughly 10 times larger. This extreme compression is why AMR was used for mobile voice recording when phone storage was measured in megabytes, not gigabytes.
You can, but it is misleading. The FLAC file will be lossless — meaning it perfectly preserves the decoded AMR audio — but the decoded AMR audio is already heavily compressed and limited. The FLAC file will be much larger than the AMR file without sounding any better. MP3 at 128 kbps is a more practical choice for AMR source material.
WhatsApp voice notes (in AMR format) can be found in your phone's WhatsApp media folder or in a Google Drive/iCloud backup. Once you locate the .amr files, convert them using FFmpeg or our audio converter. The process is the same as any other AMR file. Newer WhatsApp voice notes use Opus format instead of AMR — these are higher quality and can be converted using the same tools.
AMR was designed for 2G/3G cellular networks where bandwidth was extremely limited. Modern phones use more efficient codecs like Opus (WhatsApp, Telegram), AAC (Apple), or codec-agnostic recording formats that leverage the dramatically increased storage and bandwidth of modern devices. AMR is still used in some cellular voice networks (VoLTE uses AMR-WB), but consumer-facing applications have moved on to better alternatives.
Alex Thompson
Software engineer and content creator focused on web technologies, image optimization, and developer tooling.