Opus 1.6 Released (opus-codec.org)

Opus 1.6 has been released, introducing an experimental wideband-to-fullband speech enhancer aimed at improving speech quality in audio streams. This new Bandwidth Extension (BWE) model leverages a neural network to generate high-frequency content (8-20 kHz) from existing wideband speech (0-8 kHz), enhancing compatibility with any prior Opus version while maintaining codec integrity. Key improvements in this release also include significant enhancements to the Deep Redundancy (DRED) decoding process, which now achieves fullband quality even at low bitrates as low as 9 kb/s. The update features advanced capabilities, including support for 96-kHz audio and extensive bitrates up to 2 Mb/s, catering to high-resolution audio needs. The new 24-bit integer audio API offers a robust option for applications favoring integer operations over floating-point, enhancing compatibility with high-resolution audio pipelines. With these advancements, Opus 1.6 not only improves speech intelligibility but also redefines the boundaries of audio codec functionality, setting a precedent for future developments in audio technology within the AI/ML community.