I-SoundStream Neural Codec
SoundStream is Google's end-to-end neural audio codec that compresses speech and music to extremely low bitrates while preserving quality.
Uhlolojikelele
SoundStream is Google's end-to-end neural audio codec that compresses speech and music to extremely low bitrates while preserving quality. It matters because it beats traditional codecs like Opus at the same bitrate and powers modern generative audio models.
I-SoundStream Neural Codec ihlala ekuhambeni komsebenzi okulalelwayo-AI okuguqula inkulumo, umculo, nomsindo wokuxhumana, ukufinyeleleka, nokukhiqizwa kwemidiya.
I-Deep Dive
Yethulwe yi-Google ngo-2021, i-SoundStream iyikhodekhi ye-neural ephelele eyakhiwe ngezingcezu ezintathu eziqeqeshwe ndawonye: isifaki khodi esiguqulayo esishintsha i-waveform eluhlaza ibe ukulandelana okuhlangene kwama-vector, i-residual vector quantizer (RVQ) ehlukanisa lawo ma-vector aguqulayo futhi akha kabusha i-wavector. Iqeqeshwe ngakho kokubili ukulahlekelwa kokwakha kabusha kanye nokucwasa okuphikisayo kwesitayela se-GAN, ngakho okukhiphayo kuzwakala kungokwemvelo kunokuba nje kuvaleke ngokwezinombolo. Isici esivelele 'siyasikala' noma ukuqeqeshwa kokuyeka i-quantizer-dropout: imodeli eyodwa ingasebenza kuwo wonke ama-bitrate ukusuka cishe ku-3 ukuya ku-18 kbps ngokusebenzisa nje izendlalelo zequantizer eziningi noma ezimbalwa ekuqondeni, ngaphandle kokuqeqeshwa kabusha. Ngo-3 kbps kubikwa ukuthi idlula i-Opus ku-12 kbps ezivivinyweni zokulalela, ukuphatha inkulumo, umculo, nomsindo ojwayelekile kumodeli eyodwa engasebenza ngesikhathi sangempela ku-smartphone CPU.
I-Technical Insight
I-waveform idlula kuma-convolutions anemigqa ehlisa isampula kakhulu, ikhiqiza ukushumeka okukodwa kuhlaka ngalunye (isb. amafreyimu angu-75/isekhondi). I-RVQ bese ibhala ngekhodi ukushumeka ngakunye njengenqwaba yezinkomba ze-codebook. I-Bitrate ilingana nezinga lozimele izikhathi lapho inani lezikhawuli zezikhathi ezisebenzayo amabhithi ebhukwana ngalinye lekhodi. Ukuyeka ukuphuma kwe-Quantizer kunciphisa ngokungahleliwe isitaki se-RVQ phakathi nokuqeqeshwa, kuphoqe ama-codebook angaphambili ukuthi aphathe ulwazi olubaluleke kakhulu ukuze i-codec yonakale kahle ngamanani aphansi.
I-Mastering SoundStream Neural Codec
Ukuze wakhe ukuqonda okujulile, phatha i-SoundStream Neural Codec njengemodeli yokusebenza, hhayi isici esisodwa. Chaza imiphumela oyifunayo, cacisa ukucabanga, futhi uhlukanise lokho isistimu engakwenza ngokwethembeka kulokho okusadinga ukwahlulela kochwepheshe.
Empeleni, amaqembu aqinile asebenzisa i-SoundStream Neural Codec aphatha ikhwalithi, ukubambezeleka, kanye nemvume njengezingxenye ezibalulekile ngokulinganayo zesu lokuthunyelwa. Babhala imibandela yempumelelo ecacile, ukuhlola okuqhathaniswa nedatha engokoqobo nokugeleza komsebenzi, futhi baphindaphinde ngokusekelwe kumaphethini okuhluleka aqashiwe esikhundleni sokuwina kwebhentshimakhi yesikhathi esisodwa. Yilapho ukuqonda kwethiyori kuguquka kube amandla ahlala njalo kuwo wonke umkhiqizo, inqubomgomo, kanye nokusebenza.
Ithuthukisa ukufinyeleleka ngokuloba, ukulandisa, nezixhumi ezibonakalayo zezwi. Ngesikhathi esifanayo, ukusetshenziswa kabi kwezwi kanye nezingozi zokuzenza ongeyena ziyakhuphuka uma imvume ingekho. Indlela eqine kakhulu iwukuhlanganisa isivinini sokuhlola nesiyalo sokuphatha: qhuba abashayeli bezindiza, bamba ubufakazi, ushicilele amalogi ezinqumo, futhi ubuyekeze izivikelo ngokuqhubekayo njengoba imodeli yokuziphatha, okulindelwe ngabasebenzisi, kanye nezimfuneko zokulawula zishintsha.
I-Strategic Impact
Ithuthukisa ukufinyeleleka ngokuloba, ukulandisa, nezixhumi ezibonakalayo zezwi.
Ithuthukisa ukufinyeleleka ngokuloba, ukulandisa, nezixhumi ezibonakalayo zezwi. Ekusetshenzisweni kwekhwalithi ephezulu, lokhu kuhunyushwa emithethweni yokusebenza elinganisekayo, imingcele yobunikazi, nemikhuba yokubuyekeza ephindelelayo ukuze amaqembu akwazi ukukala ukuzethemba esikhundleni sokukala ukungaqondakali.
Amaqembu emidiya angathumela umsindo opholishiwe ngokushesha ngamabhajethi amancane.
Amaqembu emidiya angathumela umsindo opholishiwe ngokushesha ngamabhajethi amancane. Ekusetshenzisweni kwekhwalithi ephezulu, lokhu kuhunyushwa emithethweni yokusebenza elinganisekayo, imingcele yobunikazi, nemikhuba yokubuyekeza ephindelelayo ukuze amaqembu akwazi ukukala ukuzethemba esikhundleni sokukala ukungaqondakali.
Amasistimu abhekene nekhasimende angacubungula ukusebenzelana okukhulunyiwe ngesilinganiso esikhulu.
Amasistimu abhekene nekhasimende angacubungula ukusebenzelana okukhulunyiwe ngesilinganiso esikhulu. Ekusetshenzisweni kwekhwalithi ephezulu, lokhu kuhunyushwa emithethweni yokusebenza elinganisekayo, imingcele yobunikazi, nemikhuba yokubuyekeza ephindelelayo ukuze amaqembu akwazi ukukala ukuzethemba esikhundleni sokukala ukungaqondakali.
Ukuqaliswa Komhlaba Wangempela
Ukucindezela izingcingo zezwi zibe ~3 kbps kuyilapho kuzwakala kucace kunamakhodekhi efa kuma-bitrate aphezulu
Ikhiqiza amathokheni omsindo ahlukene afunza Google amamodeli akhiqizayo we-AudioLM kanye ne-MusicLM
Ukusakaza-bukhoma komsindo womkhawulokudonsa ophansi wesikhathi sangempela kumadivayisi eselula anombhalo wekhodi oku-CPU nokukhetha
Ukugcina noma ukudlulisa umculo nomsindo we-ambient ngendlela efanele ngemodeli eyodwa ephatha zonke izinhlobo zokuqukethwe
Amaphethini Okusebenzisa
I-SoundStream Neural Codec isebenza
Ukucindezela izingcingo zezwi zibe ~3 kbps kuyilapho kuzwakala kucace kunamakhodekhi efa kuma-bitrate aphezulu.
Amaqembu ngokuvamile athola imiphumela engcono uma echaza izilinganiso zekhwalithi ngaphambili, agcina indlela yokukhuphuka yomuntu yamakesi asemaphethelweni, futhi alandelele kokubili izinzuzo zokukhiqiza nezindleko zamaphutha ngokuhamba kwesikhathi.
I-SoundStream Neural Codec isebenza
Ikhiqiza amathokheni omsindo ahlukene aphakela Google amamodeli akhiqizayo we-AudioLM kanye ne-MusicLM.
Amaqembu ngokuvamile athola imiphumela engcono uma echaza izilinganiso zekhwalithi ngaphambili, agcina indlela yokukhuphuka yomuntu yamakesi asemaphethelweni, futhi alandelele kokubili izinzuzo zokukhiqiza nezindleko zamaphutha ngokuhamba kwesikhathi.
I-SoundStream Neural Codec isebenza
Ukusakaza-bukhoma komsindo womkhawulokudonsa ophansi wesikhathi sangempela kumadivayisi eselula anombhalo wekhodi oku-CPU nokukhetha.
Amaqembu ngokuvamile athola imiphumela engcono uma echaza izilinganiso zekhwalithi ngaphambili, agcina indlela yokukhuphuka yomuntu yamakesi asemaphethelweni, futhi alandelele kokubili izinzuzo zokukhiqiza nezindleko zamaphutha ngokuhamba kwesikhathi.
I-SoundStream Neural Codec isebenza
Ukugcina noma ukudlulisa umculo nomsindo we-ambient ngendlela efanele ngemodeli eyodwa ephatha zonke izinhlobo zokuqukethwe.
Amaqembu ngokuvamile athola imiphumela engcono uma echaza izilinganiso zekhwalithi ngaphambili, agcina indlela yokukhuphuka yomuntu yamakesi asemaphethelweni, futhi alandelele kokubili izinzuzo zokukhiqiza nezindleko zamaphutha ngokuhamba kwesikhathi.
Izingozi & Guardrails
Ukusetshenziswa kabi kwezwi kanye nezingozi zokuzenza ongeyena ziyanda uma imvume ingekho.
Ukunemba kungase kwehle kuzo zonke izinhlobo zokuphimisela, izilimi zesigodi, noma izindawo ezinomsindo.
Umsindo wokwenziwa ungenziwa iphutha njengenkulumo eyiqiniso ngaphandle kokulebula okucacile.
Ukuqalisa Umhlahlandlela
Thola imvume esobala yokuthwebula izwi, ukuhlanganisa, nokusebenzisa kabusha.
Phatha lokhu njengesango lobufakazi: uma imibandela ingafinyelelwa, misa ukukhishwa, vala igebe, bese unweba ukusetshenziswa.
Ikhwalithi yokuhlola kuzo zonke izipikha nezimo zangemuva.
Phatha lokhu njengesango lobufakazi: uma imibandela ingafinyelelwa, misa ukukhishwa, vala igebe, bese unweba ukusetshenziswa.
Chaza ukuthi kunini lapho umuntu kufanele abuyekeze noma agunyaze okuphumayo.
Phatha lokhu njengesango lobufakazi: uma imibandela ingafinyelelwa, misa ukukhishwa, vala igebe, bese unweba ukusetshenziswa.
Lebula umsindo wokwenziwa futhi ugcine amarekhodi atholakalayo ukuze aziphendulele.
Phatha lokhu njengesango lobufakazi: uma imibandela ingafinyelelwa, misa ukukhishwa, vala igebe, bese unweba ukusetshenziswa.
Qhubeka Uhlole
Check your understanding
Test yourself: take the SoundStream Neural Codec quiz