Audio AI JAGORA

Neural Vocoders

Vocoder na jijiyoyi wani samfuri ne wanda ke juya ƙaramin sautin sauti, yawanci mel-spectrogram, zuwa ainihin nau'in raƙuman sauti.

2 min karatuAn sabunta ta ƙarshe

Dubawa

It is the final stage that gives modern text-to-speech and voice cloning their natural, human sound.

Zurfafa nutsewa

Ƙirƙirar magana ta al'ada ta yi amfani da na'urorin sarrafa sigina waɗanda sau da yawa kamar hayaniya ko mutum-mutumi. Vocoders na ƙwanƙwasa suna koyon sake gina ɗanyen samfuran sauti daga na'urar kallo ta hanyar horarwa akan sa'o'i na rikodi na gaske. WaveNet (DeepMind, 2016) shine ci gaba, yana tsinkayar sauti ɗaya samfurin lokaci guda a samfuran 16,000+ a sakan daya, yana samar da magana mai ban mamaki amma a hankali. Samfuran daga baya sun yi ciniki da waccan ƙwanƙwasa mai ƙarfi don saurin: WaveGlow sun yi amfani da tsarar da ke gudana, Parallel WaveGAN da MelGAN sun yi amfani da cibiyoyin sadarwa na gaba, kuma HiFi-GAN ya zama sanannen ma'auni ta hanyar samar da ingantaccen sauti na 22kHz cikin sauri fiye da ainihin lokaci. A yau vocoder kusan kusan rabin na biyu na bututun mai matakai biyu, an haɗa su tare da ƙirar sauti kamar Tacotron 2 ko FastSpeech wanda ke samar da mel-spectrogram.

Fahimtar Fasaha

A mel-spectrogram yana jefar da bayanan lokaci na sauti, yana kiyaye yadda ake rarraba kuzari a cikin maɗaurin mitar na tsawon lokaci. Babban aiki na vocoder shine ƙirƙira ingantaccen tsari mai daidaituwar igiyoyin igiyar ruwa wanda girman bakansa yayi daidai da wannan shigarwar. Vocoders na tushen GAN kamar HiFi-GAN suna amfani da masu nuna wariya da yawa waɗanda ke bincika siginar a ma'auni daban-daban da lokaci-lokaci, suna tura janareta don samar da cikakkun bayanai masu kyau kamar masu jituwa da madaidaicin madaidaicin baƙar fata.

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 Neural Vocoders

Vocoders suna ƙara ƙanƙanta da sauri don su iya aiki akan wayoyi da na'urorin da aka haɗa ba tare da haɗin girgije ba. Hakanan akwai turawa ga masu amfani da vocoed na duniya waɗanda ke haɗawa ga kowane mai magana, harshe, waƙa, ko ma sautin mara magana ba tare da sake horarwa ba. Juyin layi ɗaya yana ninka muryar murya kai tsaye zuwa tsarin ƙarshe-zuwa-ƙarshe da kuma kodecs na jijiyoyi, yana ɓata layi tsakanin matakan sauti daban-daban da yanayin motsi da rage kayan tarihi da aka gabatar ta hanyar wucewa ta tsaka-tsaki.

Aiwatar da Gaskiyar Duniya

Ƙirƙirar sautin magana ta ƙarshe a cikin mataimakan rubutu-zuwa-magana kamar masu karanta allo da aikace-aikacen kewayawa

Ƙirƙirar sautin muryoyin muryoyin da aka yi amfani da su a cikin buga rubutu da kayan aikin ba da labari mai jiwuwa

Sake gina muryoyin waƙa a cikin kiɗan AI da software na murya mai kama-da-wane

Ƙaddamar da fitowar murya akan na'urar don masu magana mai wayo da na'urorin isa ga ba tare da tafiye-tafiyen uwar garken ba

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

1

Sami tabbataccen izini don ɗaukar murya, cloning, da sake amfani.

2

Gwajin ingantattun masu magana daban-daban da yanayin baya.

3

Ƙayyade lokacin da dole ne ɗan adam ya duba ko ya amince da abubuwan da aka fitar.

4

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 Neural Vocoders quiz

Instant feedback on every answer, and a shareable certificate with a verifiable ID once you pass a course.

Fara tambayoyi

Support free AI education. AI Understanding is a 501(c)(3) nonprofit — no ads, no paywall, ever. Make a donation

Jagora na gaba

HiFi-GAN dan GAN Vocoders

Tambayoyin da ake yawan yi

What is Neural Vocoders?

Vocoder na jijiyoyi wani samfuri ne wanda ke juya ƙaramin sautin sauti, yawanci mel-spectrogram, zuwa ainihin nau'in raƙuman sauti. Mataki ne na ƙarshe wanda ke ba da rubutu-zuwa-magana da murya na zamani da ke rufe sautinsu na halitta, na ɗan adam.

Menene aikin farko na vocoder na jijiyoyi?

Vocoder na jijiyoyi yana ɗaukar ƙaramin sautin ƙararrawa, yawanci mel-spectrogram, kuma yana haɗa ainihin ɗanyen sautin muryar sauti da kuke ji.

Wanne samfurin 2016 shine ci gaban da ya sa vocoders na jijiyoyi suyi sauti na halitta?

WaveNet, daga DeepMind a cikin 2016, ya haifar da samfurin sauti-by-samfuri kuma ya samar da magana mai ban mamaki na halitta, yana ƙaddamar da zamanin vocoder na jijiyoyi.

Me yasa WaveNet na asali ya yi jinkirin samar da sauti?

WaveNet yana da autoregressive: kowane samfurin sauti ya dogara da waɗanda suka gabata, don haka samar da dubun dubatar samfurori a cikin daƙiƙa yana da tsada sosai.

Wane bayani ne mel-spectrogram ke watsar da shi musamman, yana sa aikin vocoder ya fi wahala?

Mel-spectrograms suna ci gaba da girma (makamashi kowane rukunin mitar) amma sauke lokaci, don haka vocoder dole ne ya sake gina madaidaicin igiyar igiyar ruwa wanda lokaci ya yi daidai.

Ta yaya masu muryoyin tushen GAN kamar HiFi-GAN ke haɓaka gaskiya?

HiFi-GAN yana amfani da masu wariya da yawa suna duba ma'auni daban-daban da ma'auni daban-daban, suna tura janareta don samar da ingantattun jituwa da masu wucewa.