AudioLM
AudioLM bụ usoro nchọcha Google na-ebute ọdịyo ezi uche - okwu ma ọ bụ egwu piano - site n'imeso ụda dị ka asụsụ na ibu amụma ya site na token.
Nchịkọta
It matters because it showed you can produce coherent, natural-sounding audio continuations without any text transcript or musical score.
Ime miri emi
Nke Google webatara na 2022, AudioLM na-emegharị ọgbọ ọdịyo dị ka nsogbu nhazi asụsụ: ọ na-atụgharị ụdị waveform ka ọ bụrụ akara ngosi pụrụ iche wee buru amụma maka akara na-esote, dịka ụdị ederede na-ebu amụma okwu na-esote. Isi aghụghọ ya bụ usoro nhazi nke ụdị token. 'Semantic' tokens (site na ihe nlereanya dị ka w2v-BERT) na-ejide usoro ogologo oge - ụda ụda, syntax, egwu egwu - ebe akara ngosi 'acoustic' (site na SoundStream neural codec) na-ejide nkọwa dị mma dị ka njirimara ọkà okwu, timbre, na ọnọdụ ndekọ. Site na ibu amụma ihe nrịbama semantic, wee na-edozi akara ụda olu na ha, AudioLM na-ewepụta ọga n'ihu na-anọkọ ọnụ karịa ọtụtụ sekọnd ka ọ na-echekwa olu ma ọ bụ ngwa mbụ. Nyere nkeji ole na ole okwu, ọ na-aga n'ihu na-ekwu okwu n'otu olu; nyere piano, ọ na-emeziwanye n'otu ụdị ahụ.
Nghọta nka nka
A zụrụ AudioLM naanị na ọdịyo - enweghị ederede. SoundStream na-echikota ọdịyo ka ọ bụrụ akara ụda olu site na ọnụọgụ vector fọdụrụnụ, ebe w2v-BERT na-enye akara mmụta semantic siri ike. Otutu ụdị asụsụ Transformer na-ebu amụma akara n'ọkwa dị iche iche: semantic mbụ maka nhazi, emesia akara ụda olu na nke mara mma maka mwughachi ntụkwasị obi dị elu. Ihe ngbanwe nke SoundStream na-emesị tụgharịa akara ngosi ndị e buru n'amụma azụ ka ọ bụrụ usoro ebili mmiri, na-ekwenye ụda nke na-eme ka olu na okwu ọkà okwu guzosie ike.
Mmetụta atụmatụ
Nweta na iru
Ọ na-eme ka nnweta ya dịkwuo mma site na ndegharị, ịkọ akụkọ, na ntụgharị olu.
Ọnụ ego na mmefu ego
Ndị otu mgbasa ozi nwere ike ibubata ọdịyo a na-egbu maramara ngwa ngwa site na iji obere mmefu ego.
Ọsọ na ọnụ ọgụgụ
Sistemụ na-eche ihu ndị ahịa nwere ike hazie mkparịta ụka n'ọtụtụ buru ibu.
Ọdịnihu nke AudioLM
Ntụziaka dabere na token AudioLM ghọrọ ntọala maka sistemụ ndị ọzọ: Google's AudioLM echiche etinyere n'ime MusicLM maka ederede-na-egwu na SoundStorm maka ọgbọ ọsọ ọsọ, ebe mpaghara sara mbara ugbu a na-agwakọta akara semantic na ụda olu n'ofe okwu, egwu, na mmetụta ụda. Na-atụ anya ngwa ngwa, ezigbo ọgbọ, mpụta na-arụkọ ọrụ ọnụ, yana njikwa multimodal ebe ederede ma ọ bụ akara ndị ọzọ na-eduzi ụdị zụrụ azụ naanị. Otu usoro a na-emekwa ka nchegbu gbasara mwepu olu na ihe omimi nke ọdịyo.
Mmejuputa n'ezie n'ụwa
Na-aga n'ihu obere mkpirisi okwu n'otu olu ọkà okwu na ntinye ụda na-enweghị ederede
Na-emeziwanye egwu piano ọhụrụ dabara n'ụdị nke ngwa ngwa edekọtara
Na-eje ozi dị ka ọkpụkpụ azụ-ọgbọ ọdịyo maka sistemụ ederede gaa na egwu dịka MusicLM
Nnyocha n'ime nhazi okwu nke na-echekwa prosody na ndekọ egwu site na nlele
Ihe ize ndụ & okporo ụzọ nche
Iji olu eme ihe na ihe egwu mpụta ga-abawanye mgbe nkwenye na-efu.
Izi ezi nwere ike ịdaba n'ofe ụda olu, olumba ma ọ bụ gburugburu mkpọtụ.
Enwere ike imehie ọdịyo sịntetik dị ka ezigbo okwu na-enweghị akara doro anya.
Map mmejuputa
Nweta nkwenye doro anya maka ijide olu, imechi, na ijigharị.
Nwale ogo n'ofe ndị na-ekwu okwu dị iche iche yana ọnọdụ ndabere.
Kọwaa mgbe mmadụ ga-enyocha ma ọ bụ kwado nsonye.
Deba aha ọdịyo sịntetik ma debe ndekọ ihe ndekọ maka ịza ajụjụ.
Nọgide na-eme nchọpụta
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 AudioLM 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
Ntuziaka na-esote
Nkọwa okwu na edemede okwu
Ajụjụ a na-ajụkarị
What is AudioLM?
AudioLM bụ usoro nchọcha Google na-ebute ọdịyo ezi uche - okwu ma ọ bụ egwu piano - site n'imeso ụda dị ka asụsụ na ibu amụma ya site na token. Ọ dị mkpa n'ihi na ọ gosipụtara na ị nwere ike iwepụta ihe na-aga n'ihu na ọdịyo na-ada ụda na-enweghị ederede ederede ma ọ bụ akara egwu ọ bụla.
Kedu isi echiche AudioLM na-eji ewepụta ọdịyo?
AudioLM na-atụgharị ụdị ebili mmiri ka ọ bụrụ akara ngosi pụrụ iche ma na-ebu amụma ya n'usoro, na-etinye usoro akara na-esote nke ụdị asụsụ na ụda adịghị mma.
Kedu ọrụ nke token 'semantic' na AudioLM?
Ihe nrịbama nke semantic (sitere na w2v-BERT) na-egosipụta nhazi na ịdị n'otu dị elu, ebe akara ụda olu na-ejikwa nkọwa ọdịyo dị mma.
Kedu codec neural na-emepụta akara ụda olu AudioLM?
SoundStream na-eji mbelata vector quantization iji mpịakọta ọdịyo ka ọ bụrụ akara ụda ma emesia depụta akara ngosi amụma azụ azụ ka ọ bụrụ usoro ifegharị.
Kedu ihe AudioLM chọrọ ka data ọzụzụ?
Otu ihe ama ama bụ na AudioLM na-azụ naanị na ọdịyo na-enweghị akara ederede ọ bụla, usoro mmụta site na ụda ozugbo.
Kedu ihe kpatara AudioLM ji buru amụma gbasara ntụgharị okwu tupu akara ụda olu?
Ịmepụta ihe owuwu siri ike na-ebu ụzọ mee ka mmepụta ahụ dị n'otu ka oge na-aga; A na-edozi akara ụda olu na nhazi ahụ iji gbakwunye nkọwa zuru oke.