The Adaptive Multi-Rate (AMR) Codec is a speech compression standard originally developed for GSM and 3G mobile networks. Consilient Technologies provides production-ready AMR encoder/decoder libraries — both Narrowband (NB-AMR) and Wideband (WB-AMR) variants — optimized for integration into telecom infrastructure, VoIP platforms, and embedded devices. Our implementation focuses on computational efficiency, standards compliance, and ease of integration into existing audio pipelines.

NB-AMR operates at 4.75–12.2 kbps and covers a frequency range up to 3.4 kHz — suitable for legacy GSM telephony and narrowband VoIP where bandwidth is constrained. WB-AMR (also known as AMR-WB or G.722.2) operates at 6.6–23.85 kbps with audio bandwidth up to 7 kHz, delivering noticeably richer voice quality. Choose NB-AMR when you need backward compatibility with older GSM/UMTS infrastructure or when operating on very low-bandwidth links. Choose WB-AMR when your network supports it and voice quality is a priority — for example, in VoLTE deployments or HD voice-enabled VoIP systems.

AMR's defining feature is its adaptive bitrate mechanism. The codec can switch between multiple bitrate modes (eight for NB-AMR, nine for WB-AMR) on a frame-by-frame basis — every 20 ms. When the network signals congestion or increased error rates, the codec drops to a lower bitrate to reduce the data load, preserving call continuity. When conditions improve, it scales back up for better audio fidelity. This happens transparently, without interrupting the call. Consilient Technologies' implementation exposes configurable mode-set parameters so integrators can control which bitrate modes are permitted for their specific deployment.

Our AMR Codec libraries are available as portable C source code and pre-optimized builds for ARM (Cortex-A and Cortex-M series), various DSP architectures, and x86/x64 platforms. We support integration into Linux, Android, RTOS environments (FreeRTOS, ThreadX, VxWorks), and bare-metal embedded systems. This makes the codec suitable for a wide range of targets — from cloud-based media servers to battery-powered IoT voice devices.

AMR is the right choice when you need compatibility with existing 2G/3G/VoLTE mobile networks, since it remains the mandatory codec for GSM and is widely deployed in VoLTE via AMR-WB. EVS (Enhanced Voice Services) offers better audio quality at equivalent bitrates and supports fullband audio, but requires 4G/5G infrastructure and has higher computational cost. Opus is ideal for internet-based applications (WebRTC, streaming) but is not part of the 3GPP mobile standard. If your primary requirement is interoperability with mobile carrier networks, AMR is typically the most practical choice.

Common deployments include: telecom infrastructure such as media gateways and session border controllers that transcode between AMR and other formats; VoLTE and VoWiFi services where AMR-WB is the standard HD voice codec; embedded communication devices like two-way radios, in-vehicle systems, and IoT endpoints with voice capability; voice recording and storage systems that need efficient compression for large call archives; and conferencing bridges that must interwork with mobile callers.

AMR inherently operates with a 20 ms frame size and a look-ahead of 5 ms (NB) or 5 ms (WB), giving an algorithmic delay of 25 ms per direction. Our implementation minimizes additional processing overhead through optimized ACELP search routines and memory-efficient buffer management. For embedded targets, we offer fixed-point implementations that avoid costly floating-point operations, keeping CPU utilization low and latency predictable — critical for real-time voice applications.

We provide an evaluation package that includes pre-built libraries for your target platform, sample application code, API documentation, and conformance test vectors aligned with 3GPP specifications (TS 26.073 for NB-AMR, TS 26.173 for WB-AMR). Integration typically involves linking the codec library into your audio pipeline, configuring the mode set and DTX (Discontinuous Transmission) parameters, and running validation against the provided test vectors. Our engineering team offers direct integration support for enterprise customers.