Reference Formats
AMR
AMR is a narrowband speech codec standardized by ETSI and adopted by 3GPP in 1999 for GSM and early 3G mobile networks, optimized for voice at low bitrates rather than music or general audio.
Extension
.amr
Full name
Adaptive Multi-Rate
Introduced
1999
Vendor
3GPP / ETSI (developed by Ericsson, Nokia and Siemens)
What is inside the container
AMR is a minimal container built specifically around a speech codec, holding audio sampled at 8kHz across a narrow set of predefined bitrates (4.75–12.2 kbit/s for AMR-NB); a wideband variant, AMR-WB, extends this to 16kHz sampling for clearer voice. It carries essentially no metadata support — no title, artist, or descriptive fields — because it was designed as a real-time telephony transport format, not an asset-interchange or music-distribution format.
How to open it
For archives
AMR should almost never be treated as an archival master in a DAM — it's a lossy telephony codec designed for real-time voice transport at very low bitrates, not for fidelity, and it lacks any metadata infrastructure. When AMR voice memos or call recordings enter a DAM (commonly from Android devices or legacy mobile-messaging exports), the standard practice is to transcode them immediately to WAV or FLAC, attach proper descriptive metadata in the DAM's own database, and treat the original AMR only as a low-priority source reference if kept at all.
AMR grew out of a European Telecommunications Standards Institute (ETSI) effort begun in October 1997 to find a better speech codec for GSM networks, with Ericsson, Nokia and Siemens jointly developing the winning candidate through two rounds of competitive selection. ETSI selected the AMR codec in October 1998, and it was formally approved and adopted as the mandatory speech codec for GSM and, soon after, for the emerging 3GPP WCDMA standard in early 1999 — making it the audio backbone of an entire generation of mobile voice calls.
Unlike every other audio format in this list, AMR was never intended for music, podcasts, or general audio distribution — it’s purpose-built for compressed speech at extremely low bitrates (as low as 4.75 kbit/s), trading away music fidelity entirely in exchange for intelligible voice over bandwidth-constrained mobile networks. That narrow design goal is exactly why AMR still shows up today as the default format for Android voice memos and MMS voice messages, decades after its original GSM use case.
| Variant | Sample rate | Bitrate range | Use case |
|---|---|---|---|
| AMR-NB (Narrowband) | 8 kHz | 4.75–12.2 kbit/s | Traditional GSM voice calls |
| AMR-WB (Wideband) | 16 kHz | 6.6–23.85 kbit/s | HD Voice on newer mobile networks |
For a DAM, AMR is a format to convert away from immediately, not to standardize on. Voice memos, call recordings, and mobile voice-message exports arriving as .amr should be transcoded to WAV or FLAC on ingest — both to bring them into a format the DAM’s other tooling and metadata systems understand, and because AMR’s telephony-grade compression is already the lowest fidelity in the library, so there’s no preservation value in keeping it as the reference copy.
Frequently asked
Why would a DAM ever need to handle AMR files at all?
AMR is still the default recording format for voice memos on many Android devices and for MMS/voice-message attachments, so any DAM that ingests mobile-originated content — customer support call recordings, field interviews recorded on Android, forwarded voice messages — will encounter it.
Is AMR suitable for music or general audio, or only speech?
Only speech — AMR's compression is specifically tuned for the narrowband frequency range of the human voice (roughly 200–3400 Hz) and performs poorly on music, which is why it should never be used for anything other than voice recordings.
Should a DAM keep AMR as the archived version of a voice recording?
No — AMR is already the lowest-fidelity format likely to appear in a media library, so a DAM should transcode incoming AMR files to WAV or FLAC immediately on ingest rather than treat the original AMR as any kind of reference master.
Does AMR support any metadata like a recording date or speaker name?
No — the AMR container has no built-in metadata fields at all. Any descriptive information (who's speaking, when it was recorded, context) has to be captured and stored separately in the DAM's own database, since it won't travel with the file itself.
What's the difference between AMR-NB and AMR-WB?
AMR-NB (narrowband) samples at 8kHz across bitrates from 4.75 to 12.2 kbit/s, matching traditional telephone-quality voice; AMR-WB (wideband) samples at 16kHz for noticeably clearer voice, used in HD Voice calling on newer networks. Both are speech-only codecs, not suited to music.
How does a DAM's ingest pipeline convert AMR to a usable archival format?
FFmpeg, built with the libopencore-amrnb/amrwb decoding libraries, is the standard tool for converting AMR files to WAV, FLAC, or MP3 in an automated batch process — most DAM ingest pipelines handling mobile content wire this in as a default transcoding step.
Can AMR files play in a web browser for DAM preview purposes?
Not reliably — AMR has no meaningful native browser support, so a DAM should always transcode AMR to MP3 or WAV before offering an in-browser preview player, rather than serving the raw AMR file.
Is AMR royalty-free for a DAM's transcoding tools to support?
The core AMR codec's original patents date to the late 1990s standardization by Ericsson, Nokia and Siemens; most patents from that era have since expired, and open implementations like the OpenCORE AMR libraries used by FFmpeg are freely available, though a DAM building commercial products should verify current patent status if AMR handling is a significant product feature.
Sources
- ETSI selected the Adaptive Multi-Rate (AMR) codec, jointly developed by Ericsson, Nokia and Siemens, in October 1998, and it was formally adopted as the mandatory speech codec for GSM and early 3GPP WCDMA networks by early 1999. checked 2026-08-07 — Library of Congress — AMR, Adaptive Multi-Rate Speech Codec