WaveNet
WaveNet, nke DeepMind webatara na 2016, bụ netwọọkụ akwara na-emepe emepe nke na-ewepụta ụda olu otu ihe n'otu oge, na-ewepụta okwu na egwu dị egwu.
Nchịkọta
It set the modern standard for high-fidelity text-to-speech.
Ime miri emi
WaveNet bụ ihe nrụpụta akpaaka: ọ na-ebu amụma ihe nlele ọdịyo ọ bụla nwere ọnọdụ na nlele niile dị n'ihu ya, na-adịkarị na nlele 16,000 ma ọ bụ 24,000 kwa sekọnd. Isi ihe ọhụrụ ya bụ nchịkọta nke mgbagwoju anya kpatara. Causal pụtara na ihe nlereanya ahụ na-ele anya azụ na oge, na-echekwa usoro ọgbọ; dilation pụtara na oyi akwa ọ bụla na-awụpụ ọnụ ọgụgụ na-eto eto nke ukwuu, ya mere, obere nchịkọta na-ekpuchi ọtụtụ puku ihe nlele (ubi nnabata sara mbara) na-enweghị nnukwu ọnụ ahịa. N'ịbụ ndị nwere ọnọdụ na njirimara asụsụ ma ọ bụ mel-spectrogram, WaveNet na-ewepụta okwu dị n'okike karịa ụda olu concatenative na parametric nke bu ya ụzọ, na-emechi ọtụtụ oghere na ndekọ mmadụ yana ike na nsụgharị mbụ nke Google Assistant.
Nghọta nka nka
Mgbagwoju anya gbajiri agbaji bụ isi aghụghọ: site na ọnụ ọgụgụ mgbasa nke 1, 2, 4, 8, na ihe ndị ọzọ, netwọkụ naanị iri puku kwuru iri dị omimi nwere ike ịga na puku kwuru puku ihe nlele gara aga, na-ewere nkọwapụta nke ọma waveform yana ogologo usoro prosodic. Ihe mmepụta ahụ na-egosipụta uru nke ọ bụla dị ka nkesa categorical (nke mbụ ọkwa 256 site na nchịkọta iwu mu-law), yana nkeji ịgbalite gbakwunyere ihe fọdụrụ na njikọ ịwụ na-eme ka ọzụzụ nke nchịkọta a dị omimi 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 WaveNet
Original WaveNet dị nwayọ n'ihi na nlele bụ usoro. Ndị na-aga nke ọma doziri nke a: Parallel WaveNet na WaveRNN nyere aka nhazi oge, na emesịa na-asọpụta- na GAN vocoders dị ka WaveGlow na HiFi-GAN, gbakwunyere vocoders mgbasa ozi, na-akwalite àgwà na ọsọ ọsọ. Echiche nke WaveNet na-akpaghasị anya, na-agbasawanye-convolution na-ebi na sistemụ ndị a ma nwee mmetụta na-arụ ụlọ nke ọma karịa ọdịyo, na-eme ka ihe nketa ya na-emepụta ihe.
Mmejuputa n'ezie n'ụwa
Na-emepụta olu na-ada ụda sitere n'okike maka Google onye enyemaka na Google Ederede-gaa-na-okwu igwe ojii
Na-eme dị ka ụda olu na-atụgharị mel-spectrograms ka ọ bụrụ ụdị ebili mmiri na pipeline TTS dị ka Tacotron 2
Na-emekọrịta piano na egwu akụrụngwa sitere na ọdịyo raw
Ngụkọta olu maka ngwa nnweta na ịkọ akwụkwọ ọdịyo
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 WaveNet 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
Nchọpụta Deepfake Audio
Ajụjụ a na-ajụkarị
What is WaveNet?
WaveNet, nke DeepMind webatara na 2016, bụ netwọọkụ akwara na-emepe emepe nke na-ewepụta ụda olu otu ihe n'otu oge, na-ewepụta okwu na egwu dị egwu. O setịpụrụ ụkpụrụ ọgbara ọhụrụ maka ederede-gaa-okwu ziri ezi.
Kedu ka WaveNet si ewepụta ọdịyo?
WaveNet bụ autoregressive: ọ na-ebu amụma maka nlele ọdịyo ọ bụla dabere na nlele ndị bịara n'ihu ya.
Gịnị bụ isi ihe ọhụrụ convolutional na WaveNet?
Mgbagwoju anya ihe kpatara na-eme ka netwọk ahụ kpuchie puku kwuru puku ihe nlele gara aga nke ọma ka ọ na-ele anya azụ n'oge.
Kedu ihe kpatara dilation ji enyere WaveNet aka?
Ọnụego dilation na-abawanye nke ukwuu na-enye ohere nnabata sara mbara karịa ọtụtụ puku nlele nwere ọkwa ole na ole.
Kedu ihe bụ nnukwu ihe ndọghachi azụ nke WaveNet mbụ?
N'ihi na nlele ọ bụla dabere na ndị gara aga, ọgbọ na-aga n'usoro ya mere ọ dị nwayọ, na-akpali Parallel WaveNet na ndị ọzọ ga-anọchi ya.
N'ime pipeline ederede gaa na okwu, WaveNet na-arụkarị ọrụ dị ka gịnị?
WaveNet nwere ike were mel-spectrogram (dịka ọmụmaatụ, site na Tacotron 2) wee mepụta ụdị ikuku na-ada ụda ikpeazụ.