Google published Update audio codec Lyra 0.0.2 optimized to achieve maximum voice quality when using very slow communication channels. The codec was opened in early April, but was delivered in binding to the proprietary mathematical library. In version 0.0.2, this disadvantage is eliminated and an open replacement is created for the specified library – Sparse_matmul, which, as the codec itself extends under the Apache 2.0 license. From other improvements, it is possible to note the possibility of using the Bazel assembly system with the GCC multiputer and the use of this default ligament in Linux instead of Bazel + Clang.
Recall that in terms of the quality of transmitted voice data at low speeds, Lyra significantly exceeds the traditional codecs in which digital signal processing methods are used. To achieve high quality voice transmission in the conditions of limited volume of information transmitted, in addition to conventional sound compression methods and signal conversion, the speech model is used on the basis of the machine learning system, which allows you to recreate the missing information based on typical speech characteristics. The model used to generate sound is trained using several thousand hours with votes records in more than 70 languages. The performance of the proposed implementation is sufficient for encoding and decoding speech in real time on the smartphones of the average price range, with a signal transmission delay at 90 milliseconds.