Okwu NaturalSpeech na mgbasa ozi nzuzo TTS
NaturalSpeech bụ ahịrị Microsoft TTS nyocha nke na-achọ ịdị mma okwu ọkwa mmadụ, yana nsụgharị ndị ọzọ na-eji mgbasa ozi latent mepụta ụda olu bara ụba.
Nchịkọta
It shows how diffusion models, famous for images, can produce expressive, controllable audio.
Ime miri emi
Ihe izizi NaturalSpeech (2022) bụ usoro izizi a kọrọ na ọ ruru ogo ogo mmadụ na benchmark LJSpeech, nke ndị na-ege ntị na-enweghị ike ịkọwa ya nke ọma site na ndekọ ndekọ. Ọ na-eji variational autoencoder nwere nlezianya dabara tupu ya mechie oghere dị n'etiti ọzụzụ na ntinye. NaturalSpeech 2 wee nakwere usoro mgbasa ozi latent: codec audio neural na-etinye okwu n'ime vectors na-aga n'ihu, na ụdị mgbasa ozi na-amụta ịmepụta latent ndị ahụ site na ederede, na-eme ka ụda olu na-agbapụ agbaze site na obere oge. NaturalSpeech 3 webatara mgbasa ozi nke ọma, na-ekewapụta okwu n'ime àgwà ndị ekewapụrụ dị ka ọdịnaya, prosody, timbre, na nkọwa ụda, yabụ enwere ike ịhazi nke ọ bụla ma chịkwaa onwe ya maka ikwesị ntụkwasị obi dị elu na mgbanwe.
Nghọta nka nka
Mgbasa mgbagha na-arụ ọrụ site n'ịgbakwụnye mkpọtụ na nnochite anya obere okwu nke okwu na ọzụzụ netwọk iji tụgharịa mkpọtụ mkpọtụ ahụ site na nzọụkwụ. Kama ịkatọ ụdị ebili mmiri raw ma ọ bụ spectrograms zuru oke, NaturalSpeech 2 na-ekwupụta latent codec, nke dị obere ma dị mfe ịmebe. Ịkwado na ederede na olu ntụaka na-eduga ngbanwe ntụgharị ahụ, ya mere akara ngosi ikpeazụ egosipụtara n'ime okwu dabara na ọdịnaya achọrọ yana njirimara ọkà okwu.
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 Okwu Ekebe na Mgbasa Latent TTS
TTS gbadoro ụkwụ na mgbasa ozi na-arụtụ aka n'ụda ndị na-abụghị naanị ihe ebumpụta ụwa kama ọ na-arụ ọrụ nke ọma, na-ahapụ ndị ọrụ ka ha gbanwee timbre, mmetụta, na prosody dị ka ọkpụkpọ nọọrọ onwe ha. Na-atụ anya nlele ọsọ ọsọ site na ntughari na mgbasa ozi-nzọụkwụ ole na ole, cloning na-enweghị atụ siri ike site na sekọnd nke ọdịyo, yana njikọta siri ike na ụdị asụsụ buru ibu maka nnyefe-mmata gburugburu. Ọganihu ndị a na-ewusikwa mkpa maka akara mmiri na nchekwa nkwenye, ebe ọ bụ na cloning n'ezie na-ebuli ihe egwu eme ihe doro anya.
Mmejuputa n'ezie n'ụwa
Ụlọ ihe nkiri na-emepụta ihe na-emechi olu onye na-eme ihe nkiri site na obere nlele iji wepụta ihe nkiri, na-eji NaturalSpeech 2-style zero-shot cloning.
Ngwa ikpo okwu nke akwụkwọ ọdịyo na-ewepụta akụkọ ọkwa mmadụ nke ndị na-ege ntị na-agbasi mbọ ike ịmata ọdịiche na nkà olu n'ezie.
Ngwá ọrụ nnweta na-emepụtaghachi olu nke mmadụ site na ndekọ ochie maka ndị kwụsịrị ikwu okwu.
Suites imepụta ọdịnaya na-eme ka ndị editọ nwee onwe ha ịhazigharị timbre na prosody, na-eji njiri mara NaturalSpeech 3.
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 NaturalSpeech and Latent Diffusion TTS 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 NaturalSpeech and Latent Diffusion TTS?
NaturalSpeech bụ ahịrị Microsoft TTS nyocha nke na-achọ ịdị mma okwu ọkwa mmadụ, yana nsụgharị ndị ọzọ na-eji mgbasa ozi latent mepụta ụda olu bara ụba. Ọ na-egosi ka ụdị mgbasa ozi, nke a ma ama maka onyonyo, nwere ike wepụta ọdịyo na-egosipụta, na-achịkwa.
Kedu ihe dị ịrịba ama mbụ NaturalSpeech (2022) kọrọ na o nwetara?
A kọwapụtara NaturalSpeech dị ka usoro izizi ruru ogo ogo mmadụ na LJSpeech, ebe ndị na-ege ntị enweghị ike ịdabere na ọdịiche dị na ya na ezigbo okwu.
Kedu ihe ụdị mgbasa ozi latent NaturalSpeech 2 na-ewepụta n'ezie?
NaturalSpeech 2 na-agbasa n'elu codec latents, nke dị kọmpat ma dị mfe ịmegharị karịa ụdị ebili mmiri raw, wee dekọọ ha ka ọ bụrụ ọdịyo.
Kedu ka ụdị mgbasa ozi si ewepụta data na ọkwa dị elu?
Mgbasa mgbasa na-agbakwụnye mkpọtụ n'oge ọzụzụ ma mụta ịtụgharị ya, na-emepụta ihe nlele site na iji nwayọọ nwayọọ na-ajụ ụda na-enweghị usoro.
Kedu ikike isi ihe mgbasa ozi latent na-enye NaturalSpeech 2?
Site n'ịkwado ụda olu dị mkpụmkpụ, NaturalSpeech 2 nwere ike mechie olu ọkà okwu ọ na-azụbeghị nke ọma.
Gịnị bụ isi echiche nke NaturalSpeech 3's 'factorized diffusion'?
Mgbasa mgbasa ozi na-ekewapụ ọdịnaya, prosody, timbre na nkọwa ụda ka e wee nwee ike ịhazi ma jikwaa njirimara ọ bụla iche.