UGoogle upapashe uhlaziyo kwi-Lyra 0.0.2 audio codec, elungiselelwe ukufikelela kwinqanaba eliphezulu lelizwi xa usebenzisa amajelo onxibelelwano acothayo. I-codec yavulwa ekuqaleni kuka-Epreli, kodwa yanikezelwa ngokubambisana nethala leencwadi lezibalo. Kwinguqulo ye-0.0.2, le mpendulo isusiwe kwaye indawo evulekileyo yenziwe kwilayibrari ekhankanyiweyo - sparse_matmul, ethi, njenge-codec ngokwayo, isasazwe phantsi kwelayisensi ye-Apache 2.0. Olunye uphuculo lubandakanya ukukwazi ukusebenzisa inkqubo yokwakha ye-Bazel kunye nomqokeleli we-GCC kunye nokusetyenziswa kwe-bundle ngokungagqibekanga kwi-Linux endaweni ye-Bazel+Clang.
Masikhumbule ukuba ngokomgangatho wedatha yezwi edlulisiweyo ngesantya esiphantsi, i-Lyra iphezulu kakhulu kwii-codecs zendabuko ezisebenzisa iindlela zokucwangcisa umqondiso wedijithali. Ukufezekisa ukuhanjiswa kwelizwi eliphezulu kwiimeko zolwazi oluncinci olugqithisiweyo, ngaphezu kweendlela eziqhelekileyo zokunyanzeliswa komsindo kunye nokuguqulwa kwesignali, uLyra usebenzisa imodeli yentetho esekelwe kwinkqubo yokufunda ngomatshini, ekuvumela ukuba uphinde wenze ulwazi olulahlekileyo olusekelwe iimpawu zentetho eziqhelekileyo. Imodeli esetyenzisiweyo ukwenza isandi yaqeqeshwa kusetyenziswa amawaka aliqela eeyure zokurekhodwa kwelizwi kwiilwimi ezingaphezu kwama-70. Ukusebenza kophunyezo olucetywayo lwanele kwi-encoding ye-real-time ye-encoding kunye ne-decoding kwii-smartphones zexabiso eliphakathi, kunye nokulibaziseka kokuhanjiswa kwesignali ye-90 millisecond.
umthombo: opennet.ru