SoundStorm Parallel Audio Generation
SoundStorm bụ ihe atụ ọgbọ ọdịyo Google nke na-ewepụta okwu na ụda n'otu n'otu karịa otu akara n'otu oge, na-eme njikọ ọdịyo dị elu nke ukwuu ngwa ngwa.
Nchịkọta
It matters because it cuts generation latency for long clips from minutes to seconds without sacrificing fidelity.
Ime miri emi
SoundStorm, nke Google webatara na 2023, na-ewepụta ọdịyo egosipụtara dị ka akara ụda olu pụrụ iche site na codec akwara akpọrọ SoundStream. Ụdị ndị mbụ dị ka AudioLM mepụtara akara ngosi ndị a na-akpaghị aka, na-ebu amụma nke ọ bụla n'usoro, nke na-adịghị ngwa ngwa maka ogologo ọdịyo. SoundStorm na-eji usoro na-abụghị nke onwe, dabere na nkpuchi agbaziri site na ụdị ọgbọ onyonyo dịka MaskGIT. Ọ na-amalite n'ọtụtụ n'ime ihe mkpuchi mkpuchi ma na-eju ha ugboro ugboro n'ọtụtụ usoro nhazi nhazi, na-ebu amụma ọtụtụ akara n'otu oge n'otu oge. Ejiri ya na akara semantic (site na ihe nlereanya dị ka AudioLM ma ọ bụ SPEAR-TTS), ọ nwere ike ịmepụta 30 sekọnd nke mkparịta ụka sitere n'okike n'ihe dị ka ọkara nke abụọ na TPU, ihe dị ka 100 ugboro ngwa ngwa karịa ntọala autoregressive mgbe ha na-adaba àgwà ha na nkwubi okwu.
Nghọta nka nka
SoundStorm na-egosiputa ọkwa nke residual vector quantization (RVQ) sitere na SoundStream. N'oge ọzụzụ, a na-ekpuchi akara ngosi na-enweghị usoro ma ihe nlereanya ahụ na-amụta ịkọ ha. N'ịtụle ya, ọ na-eme mkpebi ngbanwe dabere na ntụkwasị obi: n'usoro nke ọ bụla, ọ na-ebu amụma akara niile kpuchiri ekpuchi, na-edobe ndị kacha nwee ntụkwasị obi, wee megharịa ndị ọzọ. Ọ na-ebu ụzọ dekọọ ọkwa RVQ siri ike, emesia nke ka mma, na-erute ọdịyo zuru oke n'usoro dị obere karịa ọgbọ token site-token.
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 SoundStorm Parallel Audio Generation
Ndozi dabere na nkpuchi na-aghọ ihe ọkọlọtọ maka ọdịyo ngwa ngwa na njikwa. Na-atụ anya na ọ ga-eme ka ndị na-akparịta ụka n'oge dị adị, nchịkọta olu ngwa ngwa, na pọdkastị ogologo oge ma ọ bụ ọgbọ audiobook ebe nkwụsịtụ mere ka ụdị autoregressive ghara ịdị irè. Ijikọta ya na nhazigharị semantic siri ike na akara mmiri ga-eme ka mkparịta ụka dịkwuo mma na nchọta. Otu echiche mmezigharị ahụ nwere ike jikọta ya na ụzọ mgbasa ozi, na-eme ka ahịrị dị n'etiti codec-token na ndị na-emepụta ọdịyo na-aga n'ihu.
Mmejuputa n'ezie n'ụwa
Ịmepụta mkparịta ụka 30 nke abụọ maka ndị enyemaka olu AI n'okpuru otu sekọnd
Na-ejikọta mkparịta ụka ọtụtụ ntụgharị na olu ọkà okwu na-agbanwe agbanwe maka imepụta ihe
Na-eme ka ederede-na-okwu dị ala dị ala na ndị ọrụ mmekọrịta ebe ụdị autoregressive na-ada
Na-emepụta ọdịyo ndekọ ogologo ụdị ọsọ ọsọ site na-ejuputa akara ụda n'otu oge
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 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
Ntuziaka na-esote
Mgbasa Ọdịda Ọdịda Stable
Ajụjụ a na-ajụkarị
What is SoundStorm Parallel Audio Generation?
SoundStorm bụ ihe atụ ọgbọ ọdịyo Google nke na-ewepụta okwu na ụda n'otu n'otu karịa otu akara n'otu oge, na-eme njikọ ọdịyo dị elu nke ukwuu ngwa ngwa. Ọ dị mkpa n'ihi na ọ na-ebelata ike ọgwụgwụ nke ọgbọ maka obere vidiyo site na nkeji ruo sekọnd n'achụghị ntụkwasị obi.
Gịnị bụ isi ihe ọhụrụ na-eme SoundStorm ngwa ngwa karịa AudioLM?
SoundStorm na-edochi ngwa ngwa autoregressive, token-site-token ọgbọ na-abụghị autoregressive myikota, na-ebu amụma ọtụtụ token n'out oge.
Kedu usoro sitere na ọgbọ onyonyo ka SoundStorm na-emegharị?
SoundStorm na-enweta atụmatụ ngbanwe dabere na ntụkwasị obi, nkpuchi na njuputa nke MaskGIT na-ewu ewu na njikọ onyonyo.
Kedu ụdị nnọchite anya SoundStorm na-ewepụta?
SoundStorm na-ebu amụma akara ngosi ụda pụrụ iche nke SoundStream codec mepụtara, bụ nke emechara dekọọ ka ọ bụrụ usoro ntụgharị.
Ogologo oge ole ka SoundStorm na-ewe iji wepụta 30 sekọnd nke ọdịyo na TPU?
SoundStorm nwere ike ịmepụta 30 sekọnd nke ọdịyo n'ime ihe dị ka 0.5 sekọnd, ihe dịka 100x ngwa ngwa karịa ntọala autoregressive.
Kedu ka SoundStorm si ekpebi akara ngosi amụma ga-edobe n'oge ngbanwe?
N'ime ntugharị ọ bụla ọ na-ejigide amụma amụma ntụkwasị obi kachasị elu ma na-ekpuchikwa ndị ejighị n'aka maka ngafe ọzọ.