Ijwi ryumuyaga Iringaniza Igisekuru cyamajwi
Ijwi ryumuyaga ni Google yerekana amajwi yerekana amajwi atanga amajwi nijwi mu buryo bubangikanye aho kuba ikimenyetso kimwe icyarimwe, bigatuma synthesis nziza yo mu rwego rwo hejuru yihuta cyane.
Incamake
It matters because it cuts generation latency for long clips from minutes to seconds without sacrificing fidelity.
Kwibira cyane
Ijwi ryumuyaga, ryatangijwe na Google mumwaka wa 2023, ritanga amajwi agaragara nkibimenyetso bya acoustic discret kuva muri codec ya neural yitwa SoundStream. Moderi yambere nka AudioLM yabyaye ibimenyetso byikora, guhanura buri kimenyetso mukurikirane, kikaba gitinda kumajwi maremare. Ijwi ryumuyaga ahubwo rikoresha uburyo butari ubwigenge, bushingiye kuri mask bwatijwe muburyo bwo kwerekana amashusho nka MaskGIT. Itangirana nibimenyetso byipfundikijwe cyane kandi byuzuye bikuzuza hejuru yintoki zintambwe, guhanura ibimenyetso byinshi icyarimwe. Bitunganijwe ku bimenyetso bisobanutse (uhereye ku cyitegererezo nka AudioLM cyangwa SPEAR-TTS), irashobora guhuza amasegonda 30 y'ibiganiro bisanzwe mu gice cya kabiri cy'isegonda kuri TPU, byikubye inshuro 100 byihuta kurenza ibice byigenga mugihe bihuye nubwiza bwabo hamwe no kuvuga neza.
Ubushishozi
Ijwi ryumuyaga ryerekana urwego rwibisigisigi bya vector bisigaye (RVQ) kuva kuri SoundStream. Mugihe cyamahugurwa, ibimenyetso bidasanzwe byapimwe kandi icyitegererezo cyiga kubihanura. Mugusuzuma ikora ikizere gishingiye kuburinganire bwa decoding: muri buri itera iteganya ibimenyetso byose bipfutse mu maso, bikomeza ibyiringiro cyane, kandi byongeye guhisha ibisigaye. Isohora urwego ruto rwa RVQ mbere, hanyuma nziza, igera kumajwi yuzuye murwego ruto cyane ugereranije na token-by-token generation.
Ingaruka z'Ingamba
Kugera no kugera
Itezimbere kugerwaho binyuze mu kwandukura, kuvuga, no guhuza amajwi.
Igiciro na bije
Amatsinda yibitangazamakuru arashobora kohereza amajwi yihuse hamwe na bije nto.
Umuvuduko n'igipimo
Sisitemu ireba abakiriya irashobora gutunganya imikoranire ivugwa murwego runini.
Ejo hazaza ha Ijwi ryumuyaga Iringaniza amajwi
Kuringaniza mask ishingiye kuri decoding ihinduka igikoresho gisanzwe cyamajwi yihuta. Witegereze imbaraga zigihe-nyacyo cyo kuganira, guhuza amajwi ako kanya, hamwe na podcast ndende cyangwa ibisekuruza byamajwi aho ubukererwe bwigeze gukora moderi ya autoregressive idakwiye. Kubihuza hamwe nubusobanuro bukomeye hamwe nibisobanuro byamazi bizamura ibiganiro byukuri no gukurikirana. Igitekerezo kimwe cyo gutondeka-kunonosora igitekerezo gishobora guhuzwa nuburyo bwo gukwirakwiza, bigahindura umurongo uri hagati ya codec-token na moteri ikomeza-amajwi.
Gushyira mu bikorwa Isi
Gukora 30-isegonda ivugwa kubiganiro bifasha amajwi AI munsi yisegonda
Gukomatanya ibiganiro byinshi-hamwe n'amajwi ahoraho avuga kuri prototyping
Guha imbaraga-ubukererwe bwinyandiko-ku-mvugo mubikorwa bikorana aho moderi ya autoregressive itinda
Gukora amajwi maremare yavuzwe byihuse wuzuza ibimenyetso bya acoustic murwego rumwe
Ingaruka & Kurinda
Gukoresha nabi amajwi no kwigira ibyago byiyongera mugihe uruhushya rubuze.
Ukuri kurashobora kugabanuka hejuru yimvugo, imvugo, cyangwa urusaku rwibidukikije.
Amajwi yubukorikori arashobora kwibeshya kumvugo yukuri nta kirango gisobanutse.
Igishushanyo mbonera
Shaka uruhushya rusobanutse rwo gufata amajwi, gukoroniza, no gukoresha.
Ikizamini cyiza mubiganiro bitandukanye hamwe nuburyo bwimbere.
Sobanura igihe umuntu agomba gusuzuma cyangwa kwemeza ibisubizo.
Andika amajwi yubukorikori kandi ugumane inyandiko zerekana kubazwa.
Komeza Ubushakashatsi
Free newsletter
Get the daily AI briefing
Three verified AI stories every weekday morning, written in plain English. Free forever, no ads.
One email each weekday. Unsubscribe in one click. We never sell or share your address.
Test yourself
Take the SoundStorm Parallel Audio Generation quiz
Instant feedback on every answer, and a shareable certificate with a verifiable ID once you pass a course.
Support free AI education. AI Understanding is a 501(c)(3) nonprofit — no ads, no paywall, ever. Make a donation
Ubuyobozi bukurikira
Ijwi rihamye Gutandukana
Ibibazo bikunze kubazwa
What is SoundStorm Parallel Audio Generation?
Ijwi ryumuyaga ni Google yerekana amajwi yerekana amajwi atanga amajwi nijwi mu buryo bubangikanye aho kuba ikimenyetso kimwe icyarimwe, bigatuma synthesis nziza yo mu rwego rwo hejuru yihuta cyane. Nibyingenzi kuko bigabanya ubukererwe bwibisekuru kumashusho maremare kuva kumunota kugeza kumasegonda utitangiye ubudahemuka.
Ni ubuhe buryo bw'ingenzi butuma SoundStorm yihuta kurusha AudioLM?
IjwiStorm risimbuza buhoro buhoro autoregressive, token-by-token generation hamwe na autoregressive parallel decoding, guhanura ibimenyetso byinshi icyarimwe.
Ni ubuhe buhanga buva mu mashusho yerekana amajwi ahuza?
Ijwi ryumuyaga ryagurijwe ibyiringiro, mask-na-kuzuza decoding stratégale ikunzwe na MaskGIT muguhuza amashusho.
Ni ubuhe buryo bwo kwerekana amajwi yerekana?
Ijwi ryumuhanuzi rihanura ibimenyetso bya acoustique byakozwe na codec ya SoundStream, nyuma bigahinduka muburyo bwa flake.
Hafi ya SoundStorm ifata igihe kingana iki kugirango itange amasegonda 30 y amajwi kuri TPU?
Ijwi ryumuyaga rishobora guhuza amasegonda 30 y amajwi mumasegonda hafi 0.5, hafi 100x yihuta kuruta ibyingenzi.
Nigute SoundStorm ihitamo ibimenyetso byahanuwe kubika mugihe cya decoding?
Muri buri iterisiyo igumana ibyiringiro byayo byinshi byerekana ibimenyetso kandi ikongera ikabipfukirana ibitazwi neza kuri pass ikurikira.