WaveNet
WaveNet, wanda DeepMind ya gabatar a cikin 2016, wata hanyar sadarwa ce ta ci gaba wacce ke haifar da ingantaccen sautin samfuri ɗaya a lokaci guda, yana samar da magana mai ban sha'awa da kiɗa.
Dubawa
It set the modern standard for high-fidelity text-to-speech.
Zurfafa nutsewa
WaveNet sigar ƙirar ƙira ce ta autoregressive: tana tsinkaya kowane samfurin sauti da aka sharadi akan duk samfuran da ke gabansa, yawanci a samfuran 16,000 ko 24,000 a sakan daya. Babban bidi'a ita ce tari na rikice-rikice masu lalacewa. Causal yana nufin samfurin kawai yana kallon baya a cikin lokaci, yana kiyaye tsarin tsara; dilation yana nufin kowane Layer ya tsallake adadin samfurori masu girma da yawa, don haka matsakaicin tari ya rufe dubban samfurori (filin karɓa mai faɗi) ba tare da tsada mai yawa ba. An daidaita shi akan fasalulluka na harshe ko mel-spectrogram, WaveNet yana samar da magana mafi dabi'a fiye da madaidaitan vocoders waɗanda suka gabace shi, yana rufe yawancin gibin ga rikodin ɗan adam da kuma ƙarfafa sigar farko na Google Mataimakin.
Fahimtar Fasaha
Dillated convolutions su ne mabuɗin dabara: tare da ƙimar dilation na 1, 2, 4, 8, da sauransu, hanyar sadarwa kawai dubun-dubatar zurfi za ta iya halartar dubban samfuran da suka gabata, suna ɗaukar cikakkun cikakkun bayanai na igiyar ruwa da tsayin daka. Samfuran samfuran kowane ƙima na samfurin azaman rarrabuwar kayyade (asali matakan 256 ta hanyar mu-law companding), da raka'o'in kunnawa gated tare da ragowar haɗin kai da tsallake-tsallake suna tabbatar da horon wannan tulin mai zurfi.
Dabarun Tasiri
Shiga ku isa
Yana inganta samun dama ta hanyar rubutu, ba da labari, da mu'amalar murya.
Kudin da kasafin kuɗi
Ƙungiyoyin kafofin watsa labaru na iya jigilar sauti mai gogewa cikin sauri tare da ƙaramin kasafin kuɗi.
Gudu da sikelin
Tsarin fuskantar abokin ciniki na iya aiwatar da hulɗar magana a mafi girman ma'auni.
Makomar WaveNet
Asalin WaveNet ya kasance yana jinkirin saboda samfurin yana kan layi. Masu nasara sun gyara wannan: Parallel WaveNet da WaveRNN sun ba da damar haɗin kai na lokaci-lokaci, kuma daga baya kwarara- da masu muryoyin tushen GAN kamar WaveGlow da HiFi-GAN, da masu watsa vocoders, haɓaka inganci da sauri gaba. WaveNet's autoregressive ra'ayoyin ra'ayoyin juyin juya hali suna rayuwa a cikin waɗannan tsarin kuma sun yi tasiri ga gine-gine fiye da sauti, suna tabbatar da gadonsa a cikin ƙirar ƙira.
Aiwatar da Gaskiyar Duniya
Samar da sautunan sauti na dabi'a don Google Mataimakin da Google Rubutu-zuwa-Magana ta gajimare
Yin aiki azaman vocoder na jijiyoyi wanda ke juya mel-spectrograms zuwa nau'ikan raƙuman ruwa a cikin bututun TTS kamar Tacotron 2
Haɓaka ainihin piano da kiɗan kayan aiki daga ɗanyen sauti
Haɗin murya don kayan aikin samun dama da ba da labari mai jiwuwa
Hatsari & Tsare-tsare
Rashin amfani da murya da haɗarin kwaikwaya yana ƙaruwa lokacin da aka rasa izini.
Daidaituwa na iya faɗuwa cikin lafuzza, yaruka, ko mahalli masu hayaniya.
Ana iya kuskuren sauti na roba don ingantacciyar magana ba tare da bayyananniyar lakabi ba.
Taswirar Hanya
Sami tabbataccen izini don ɗaukar murya, cloning, da sake amfani.
Gwajin ingantattun masu magana daban-daban da yanayin baya.
Ƙayyade lokacin da dole ne ɗan adam ya duba ko ya amince da abubuwan da aka fitar.
Yi lakabin sauti na roba da kuma adana bayanan da aka tabbatar don yin lissafi.
Ci gaba da Bincike
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
Jagora na gaba
Gano Deepfake Audio
Tambayoyin da ake yawan yi
What is WaveNet?
WaveNet, wanda DeepMind ya gabatar a cikin 2016, wata hanyar sadarwa ce ta ci gaba wacce ke haifar da ingantaccen sautin samfuri ɗaya a lokaci guda, yana samar da magana mai ban sha'awa da kiɗa. Ya kafa ma'auni na zamani don ingantaccen rubutu-zuwa-magana.
Ta yaya WaveNet ke samar da sauti?
WaveNet ne autoregressive: yana tsinkaya kowane samfurin sauti dangane da samfuran da suka zo gabanin sa.
Menene mabuɗin ƙirƙira juyin juya hali a cikin WaveNet?
Rarraba rikice-rikicen sanadi suna barin hanyar sadarwar ta rufe dubunnan samfuran da suka gabata da kyau yayin kallon baya kawai cikin lokaci.
Me yasa dilation ke taimakawa WaveNet?
Haɓaka ƙimar haɓakawa mai fa'ida yana ba da fa'ida mai fa'ida mai karɓa sama da dubunnan samfurori tare da ɗan yadudduka kaɗan.
Menene babban koma baya mai amfani na ainihin WaveNet?
Saboda kowane samfurin ya dogara da waɗanda suka gabata, tsararru tana kan layi don haka a hankali, yana motsa Parallel WaveNet da sauran magada.
A cikin bututun rubutu-zuwa-magana, WaveNet yakan zama me?
WaveNet na iya ɗaukar mel-spectrogram (misali, daga Tacotron 2) kuma ya samar da yanayin sautin sauti na ƙarshe.