AMR MIME Type

What does audio/amr mean?

The audio/amr media type identifies speech encoded with Adaptive Multi-Rate narrowband, usually called AMR or AMR-NB. The codec was designed for cellular voice rather than music. It works at an 8 kHz sample rate, processes speech in 20 millisecond frames, and can switch among eight coding modes from 4.75 to 12.2 kbit/s.¹

An .amr file normally contains a small header followed by consecutive encoded speech frames. The same codec can also appear inside a 3GP container, but that container uses audio/3gpp or video/3gpp, not audio/amr.¹

How to identify an .amr file

RFC 4867 defines exact headers for the standalone storage format:¹

The newline byte 0A is part of each magic number. Checking only the visible #!AMR characters can confuse a narrowband file with AMR-WB or a multi-channel variant.

After the header, each stored speech frame begins with a one-byte frame header. Its frame-type field identifies the coding mode, which in turn determines how many speech bits follow. That is why changing the extension cannot turn AMR into another audio format.

The IANA registration assigns .amr and the case variant .AMR to stored AMR data. Filename case does not create a different format.¹

Where AMR came from

ETSI developed and standardized AMR for GSM cellular systems. Selection testing took place in 1998, and 3GPP approved AMR as its mandatory narrowband speech codec in early 1999.² The codec combined eight operating modes so a mobile network could trade speech bitrate against channel protection as radio conditions changed.

The original standards targeted circuit-switched mobile radio, but the codec's mode flexibility also made it useful for packet voice, streaming, voice messages, and compact recordings. RFC 4867 standardized its RTP payload and standalone file storage formats in 2007, replacing the earlier RFC 3267 definition.¹

AMR remains a practical interchange format when working with older mobile recordings, voicemail systems, and telephony archives. Its low bitrate is an advantage for speech under tight bandwidth limits, but the narrow 8 kHz sampling is a poor fit for music or high-fidelity production.

AMR, AMR-WB, and 3GP are not the same thing

These names describe related but incompatible layers:

AMR-WB improves speech bandwidth and supports higher coding rates, but a narrowband decoder cannot interpret it merely because both names contain AMR. A 3GP file adds container structures useful for seeking, synchronization, and media tracks. Renaming message.3gp to message.amr does not remove that container.¹

How to open an audio/amr file

Native web playback is unreliable. MDN currently reports no AMR support in Chrome, Firefox, or Opera and leaves Edge and Safari support uncertain. Chromium's own supported codec and container list also omits AMR.³⁴

For local playback or inspection:

• VLC lists 3GPP AMR among its supported audio codecs.⁵
• FFmpeg supports both reading and writing the 3GPP AMR file format, and includes native AMR-NB and AMR-WB decoders.⁶
• Android's media platform supports AMR-NB encoding and decoding in .amr and .3gp files. Device applications can therefore be more capable than a desktop web browser.⁷

Support in Poly

Poly has an explicit audio/amr type and maps .amr filenames to it. Recognition is currently ahead of processing support, however. Poly's audio loader uses Symphonia with its complete codec and container feature set, and that set has no AMR codec or standalone AMR demuxer. As a result, a raw .amr file cannot complete the normal metadata, transcription, or audio-embedding pipeline.

The practical behavior is:

• Poly stores and syncs the file and indexes its filename and ordinary file properties.
• It routes the file to the audio player, but that player receives the original AMR data. Playback therefore depends on a browser decoder that is usually absent.
• It does not currently extract duration or tags, generate a transcript, or create audio-similarity segments from raw AMR.
• The agent can locate and manage the file by its metadata, but the direct multimodal-audio path explicitly rejects audio types outside its supported list, which does not include audio/amr.

This is narrower than Poly's support for common audio formats. Converting a working copy to WAV, FLAC, MP3, AAC, or Opus before upload enables the corresponding audio pipeline. Keep the original .amr when provenance matters.

Converting AMR safely

Choose an output based on what happens next:

• WAV or FLAC preserves the decoded samples without another lossy encoding. It cannot restore frequencies or detail discarded by AMR.
• Opus is useful for compact speech delivery in modern applications and browsers.
• AAC in M4A or MP3 is convenient when compatibility with consumer players matters most.

FFmpeg can decode AMR and create a more widely supported copy:⁶

# Decode to uncompressed PCM for editing or preservation work.
ffmpeg -i message.amr message.wav

# Create a compact Opus copy for web delivery.
ffmpeg -i message.amr -c:a libopus -b:a 24k message.opus

The first conversion avoids an additional lossy generation, although the output will be much larger. The second is lossy. Repeated conversion among AMR, MP3, AAC, and Opus can accumulate audible artifacts, so always retain the original when it may be needed later.

Security and privacy considerations

AMR carries recorded speech and provides no built-in confidentiality, authentication, or integrity protection. RFC 4867 requires external protection for secure transport and recommends SRTP where appropriate for RTP sessions.¹ A standalone file likewise needs access controls or encrypted storage when its contents are sensitive.

Treat unexpected voice messages as untrusted binary input. Use maintained decoders, and avoid relying on the extension alone when accepting uploads. Converting a file does not make a confidential recording safe to publish; listen to it and review its context before sharing.